Teradata Corp (NYSE: TDC) — Cash Yield Awaits an ARR Proof Point

Published: 2026-09-14 · Verdict: Accumulate · Entry price: $25 · Price target: $35 · Research confidence: High (84%)

Executive conclusion

Analyst Take

Teradata is best underwritten as a financially repaired, cash-generative installed-base franchise—not as a demonstrated AI growth platform. At the verified September 11, 2026 closing price of $27.84, 93.4 million issued shares imply approximately $2.60 billion of equity value. June cash was $414 million, the company had repaid its $450 million term loan, and finance-lease liabilities were $91 million. Including those leases as debt produces enterprise value near $2.28 billion. Against management’s 2026 guidance of $2.65-$2.73 in adjusted EPS and $330-$350 million of adjusted free cash flow, the stock trades at about 10.3 times midpoint adjusted EPS and 7.6 times midpoint adjusted free cash flow, a 13.1% stated yield. Selective accumulation below $30 is defensible, with a preferred entry near $25 and a 12-month base value of $35. Those are analyst estimates, not reported facts, and they require delivery of a heavily second-half- and Q4-weighted ARR plan. [S2] [S3] [S9]

The headline yield is not the economic yield. Teradata’s free-cash-flow definition deducts cash purchases of property and equipment and capitalized software, but equipment acquired through finance leases bypasses that capex line and principal repayment is classified as financing cash flow. Finance-lease principal was $82 million in 2023, $71 million in 2024, $69 million in 2025, and $32 million in the first half of 2026. The leases primarily finance equipment used to deliver customer services, so ignoring them makes a hardware-backed portion of recurring revenue look costless. Deducting an estimated $50-$70 million of annual replacement or lease funding from the $340 million guidance midpoint gives a provisional owner-earnings range of roughly $270-$290 million, or about a 10.4%-11.2% equity yield at the current market capitalization. The range is deliberately broad because principal repayment is not identical to maintenance capital. [S1] [S2]

Stock compensation is the second qualification. It was $112 million in 2025 and $62 million in the first half of 2026, and management excludes it from adjusted operating profit and EPS. Repurchases have nevertheless reduced shares outstanding, so deducting every dollar of stock compensation while also using the lower share count would double-count part of the cost. The right test is whether diluted shares keep falling after awards and how much cash is required to produce that decline. In 2025, $140 million of repurchases reduced year-end shares outstanding by only about 2.6 million, illustrating that a meaningful part of capital return offsets employee issuance. [S1] [S2]

The operating evidence supports stabilization, not a growth inflection. Q2 Total ARR was $1.509 billion, up 1% reported and 2% in constant currency; Public Cloud ARR was $686 million, up 8% reported and 9% in constant currency. Recurring revenue grew 3%, GAAP operating margin reached 11.7%, and non-GAAP operating margin reached 21.5%. But cloud ARR was also $686 million in Q1, Q3 revenue is guided down, and management expects most full-year ARR growth in Q4. The midpoint of 2%-4% year-end Total ARR growth implies approximately $1.568 billion, requiring about $59 million of growth from Q2. New platform products did not materially contribute in the first half. [S2] [S3] [S5] [S6] [S21]

Forecast risk is higher than the recommendation initially appears to allow. In February 2024, management guided constant-currency Total ARR growth of 4%-8% for that year; actual 2024 performance was a 4% decline. The company subsequently restructured, reduced its workforce, changed strategic emphasis, and reset outstanding long-term compensation targets. Upfront license accounting and elongated customer decisions explain some volatility, but they do not erase a ten-point miss at the midpoint. [S4] [S22] [S23]

The central variant perception is therefore narrow. Teradata does not need to grow like Snowflake or MongoDB to create value at this price. It needs to retain the installed base, produce low-single-digit Total ARR growth, sustain post-restructuring margins, and retire shares without re-levering. Hybrid and sovereign deployment may matter for regulated or latency-sensitive workloads, while OneLake and Iceberg interoperability may keep Teradata relevant when data resides elsewhere. The strongest counter-case is that these features merely slow erosion: revenue declined from $1.917 billion in 2021 to $1.663 billion in 2025, cloud net expansion fell to 108%, and faster-growing data platforms demonstrate that spending is available elsewhere. [S1] [S8] [S13] [S14] [S18]

Investment conviction is moderate. Filing evidence, cash-flow reconciliation, and balance-sheet evidence are strong. Evidence is materially weaker for total-company retention, new-logo wins, partner-sourced ARR, cloud unit economics, AI pricing, and competitive displacement. The near-term decision sequence is Q3 Total ARR, the remaining Q4 requirement, sequential cloud ARR, recurring gross margin, lease additions and principal, adjusted free cash flow, and diluted shares. Conviction would rise after two consecutive quarters of at least 3% Total ARR growth, renewed sequential cloud expansion, attributable paid AI or Factory workloads, and lease-adjusted owner earnings above $300 million. It would fall if Total ARR contracts, cloud ARR remains near $686 million through year-end, adjusted free cash flow falls below $300 million, or compensation targets are reset again after underperformance.

Verdict: Accumulate selectively because repaired liquidity and normalized cash yield create valuation support, but the growth thesis remains unverified and management’s forecasting record warrants a discount.

Stock Price Action — Five-Year Event Map

Company Financials’ split-adjusted daily series shows a five-year closing high of $58.89 on October 25, 2021 and a closing low of $19.20 on April 7, 2025. The corresponding intraday extremes were $59.38 on October 22, 2021 and $18.43 on April 7, 2025. The distinction matters: the intraday extremes should not be described as closing prices. The September 11, 2026 close of $27.84 was 52.7% below the five-year closing high and 45.0% above the closing low. Over the latest 52 weeks, the closing range was approximately $20.42-$37.88, placing the current price about 42.5% of the way from the closing low to the high. Price observations are facts; the event explanations below are interpretations tied to contemporaneous disclosures. [S9]

• October 2021 — cycle high. Teradata was then receiving a much higher multiple for subscription conversion and rapid cloud expansion. The subsequent operating record did not validate that valuation: revenue fell from $1.917 billion in 2021 to $1.663 billion in 2025, and remaining performance obligations declined from more than $2.6 billion to $2.182 billion. The price high is factual; attributing it to cloud-transition optimism is an inference from the company’s contemporary strategy and later derating. [S1] [S18]

• February 13, 2024 — 21.7% one-day decline, from $48.79 to $38.22. The move followed full-year 2023 results and management’s initial 2024 outlook. Importantly, the release itself guided constant-currency Total ARR growth of 4%-8%, so it is too simple to say the decline was caused by a disclosed ARR contraction. The market evidently judged the outlook, revenue mix, or implied conversion less favorably than the headline guidance. Actual 2024 Total ARR later declined 4% in constant currency, vindicating skepticism about the plan but not proving the exact cause of the event-day move. [S9] [S23]

• August 6, 2024 — 14.1% decline, from $29.15 to $25.05. The move followed a material guidance reduction. Q2 Total ARR was down 4% reported and 3% in constant currency, and full-year constant-currency Total ARR guidance was cut to a decline of 2%-4%. The company also announced a restructuring affecting roughly 9%-10% of employees and targeting $75-$80 million of annualized operating-expense savings. Attribution confidence is high because the price gap immediately followed the filing. [S9] [S22]

• April 7, 2025 — five-year closing low of $19.20 and intraday low of $18.43. There was no identified same-day Teradata operating disclosure establishing a company-specific cause. The low occurred after repeated guidance disappointments and 2024 ARR contraction, but broader market conditions may also have contributed. Treating the date as proof of a fundamental trough would be hindsight. [S9] [S12]

• November 5, 2025 — 32.6% one-day increase, from $20.71 to $27.46. Q3 results showed Total ARR up 1% reported and flat in constant currency, Public Cloud ARR up 11%, adjusted free cash flow of $88 million, and maintained full-year guidance. The operating improvement was modest relative to the price response, suggesting positioning and very low expectations amplified the move. Year-end Total ARR subsequently reached $1.522 billion, up 3% reported but only 1% in constant currency. [S1] [S9] [S24]

• February 11, 2026 — 29.6% increase, from $29.23 to $37.88. The move followed full-year 2025 results and announcement of the Lynrock Lake cooperation agreement. The SAP settlement was reached on February 19 and therefore cannot explain the February 11 jump. The filing showed improving Total ARR and cloud ARR, while the governance agreement committed the board to refreshment. Both plausibly contributed. [S9] [S11] [S25]

• August 5, 2026 — 23.7% decline, from $34.39 to $26.23. Q2 margins and cash generation improved, but management guided Q3 recurring revenue down 4%-2% and total revenue down 6%-4%, while maintaining a Q4-heavy ARR plan. Cloud ARR was sequentially flat. The move is most plausibly interpreted as a reset of growth and guidance-confidence expectations, not a liquidity event. [S2] [S3] [S5] [S9]

The current price is below the average price paid for 2024 repurchases and first-half 2026 repurchases, but above the 2025 repurchase average. That comparison does not prove undervaluation; it shows that historical buyback timing was inconsistent. The August 2026 gap reduced the multiple even as adjusted EPS and adjusted free-cash-flow guidance increased, creating the present valuation tension: better near-term cash metrics but less confidence that ARR growth is durable. [S1] [S2]

Verdict: The five-year map supports a low-expectations setup, but it also documents repeated forecast resets. The market’s skepticism is not merely momentum; it has a fundamental history.

Business Overview

Teradata sells database, analytics, and data-management infrastructure to large enterprises and public-sector organizations. Its historical core is a massively parallel processing database optimized for large, complex analytical workloads and high query concurrency. The current portfolio spans public cloud, customer premises, private or sovereign environments, and hybrid combinations. Customers use the platform to consolidate and govern data, execute recurring analytical queries, support operational decisions, run machine-learning workflows, and increasingly provide context to generative-AI applications. The Autonomous Knowledge Platform, AI Studio, Tera-branded agents and interfaces, elastic compute, Teradata Factory, and Microsoft OneLake integration extend that architecture. These capabilities are company descriptions; their economic impact must be established through paid adoption, retention, gross profit, and cash returns. [S1] [S7] [S8] [S19]

The business is understandable at the economic level: retain large installed workloads, migrate or expand them across deployment types, price software and compute above infrastructure cost, control customer-equipment investment, and keep research and selling expense below recurring gross profit. Product vocabulary changes faster than those drivers. The major analytical unknowns are how much ARR represents migration rather than incremental spending, how total retention differs by deployment, and whether AI features create additional consideration rather than consume capacity customers already purchased. [S1] [S2] [S5]

Revenue model and recognition

Teradata monetizes customers through term software licenses, public-cloud consumption or capacity arrangements, maintenance and upgrade rights, hardware rental, and consulting. Total ARR is the annual value of active, contractually binding recurring arrangements across public cloud, recurring AI services, subscriptions, hardware rental, maintenance, and upgrade rights. Public Cloud ARR is the subset deployed in public-cloud environments. ARR is a management operating metric rather than GAAP revenue, and it does not remove timing differences. A term license that conveys a right to use software may be recognized upfront, support is generally recognized over time, and cloud revenue may depend on capacity or consumption. [S1] [S2]

In 2025, recurring revenue was $1.445 billion, 86.9% of total revenue. Consulting contributed $201 million, or 12.1%, while perpetual software licenses, hardware, and other revenue contributed only $17 million. The high recurring mix improves visibility, but it should not be confused with an automatically expanding SaaS annuity. Recurring contracts can be as short as one month, although many are multi-year. Renewals, downsizing, consumption, cancellations, deployment migration, and upfront license accounting can all cause quarterly volatility. [S1]

Revenue is contract-heavy but not annuity-like: 86.9% of 2025 revenue was recurring, yet renewals, cloud consumption, on-premises license timing, migrations, rentals, and cancellations can still produce material volatility. Management said unusually high first-half 2026 recognition of annual on-premises licenses created difficult second-half comparisons without an equivalent change in ARR. This is arithmetically plausible, but investors should prioritize Total ARR, retention, and cash collection over one quarter’s revenue label. [S1] [S5]

Remaining performance obligations were $2.182 billion at year-end 2025, with $1.473 billion expected to be recognized within one year. Deferred revenue was approximately $544 million including current and non-current balances. These amounts provide visibility and favorable customer financing, but neither is equivalent to high-margin backlog: the obligations include future service delivery, and the mix can include cloud infrastructure and hardware-backed revenue. [S1]

Segment and unit economics

Teradata reports Product Sales and Consulting Services as operating segments. Product Sales includes recurring revenue plus perpetual software, hardware, and other sales. In 2025, Product Sales produced $1.462 billion of revenue and $998 million of segment gross profit, a 68.3% segment margin. Consulting generated $201 million of revenue and only $12 million of segment gross profit, about 6.0%, before excluded stock compensation and reorganization costs. On a GAAP presentation, consulting gross profit was approximately zero in 2025 and negative in Q2 2026. [S1] [S2]

Consulting therefore has strategic rather than demonstrated standalone economic value. It can help implement difficult workloads, expose Teradata to customer architecture decisions, and enable product adoption. Conversely, low or negative gross margin may signal poor utilization, price pressure, partner transition costs, or weakening influence over implementations. Management’s aspiration for low-double-digit consulting margin is not evidence that the business has reached it. A partner-led model could improve capital efficiency, but only if partners originate or expand profitable product ARR. [S2] [S5]

The recurring line itself is not pure software. Rental revenue, including hardware maintenance, was $212 million in 2025. Teradata owns or leases equipment used to provide some customer services and often finances that equipment to align payments with customer receipts. This improves contractual matching, but it means a portion of recurring revenue requires physical capital and supplier execution. A dollar of public-cloud ARR, subscription software ARR, maintenance ARR, and Factory-related rental ARR can have different gross margins and reinvestment needs. [S1]

Geography and customers

United States revenue was $829 million in 2025 and international revenue was $834 million, an almost even split. No individual customer represented more than 10% of revenue. Geographic diversity limits dependence on one economy or customer, but it adds currency, tax, sanctions, export-control, data-residency, and political risks. At June 2026, $359 million of the company’s $414 million of cash was held outside the United States. Management expects repatriation under the territorial tax system, but the location remains relevant to deployment timing and tax leakage. [S1] [S2]

Large customers are economically attractive because the platform can support numerous mission-critical workloads. They also create concentrated procurement power even when no single customer exceeds the disclosure threshold. A few large renewals can change quarterly ARR, and customers can stage purchases across platforms. The absence of a greater-than-10% customer reduces single-name risk but does not eliminate renewal concentration.

Deployment economics

Public cloud improves relevance, flexibility, and contractual duration but introduces third-party infrastructure expense. Recurring gross margin fell from 70.2% in 2024 to 68.0% in 2025 principally because public cloud became a larger portion of recurring revenue, partly offset by improved public-cloud margins. Q2 2026 GAAP recurring gross margin was 66.9%, up from 66.4% a year earlier; management’s non-GAAP recurring margin was 67.8%. The difference reflects excluded costs and should not be blurred. [S1] [S2] [S3]

On-premises term software may carry superior near-term accounting margin because intellectual property is delivered upfront. Hardware rental requires equipment, maintenance, and lease financing. Public cloud requires hosting capacity purchased from hyperscalers. The company therefore faces a genuine mix trade-off: cloud and hybrid deployment may improve product durability while lowering consolidated gross margin unless infrastructure purchasing and utilization improve.

Unrecognized assets and hidden obligations

The balance sheet recognized approximately $397 million of goodwill at June 2026, but the more important economic assets are mostly absent: decades of optimized database code, 534 reported U.S. patents, trade secrets, industry data models, installed schemas and queries, governance rules, customer-specific workflow integration, and employee expertise. Research and development is expensed as incurred; $280 million of 2025 spending therefore reduced earnings without creating a reported research asset. [S1] [S2]

The principal unrecognized assets are the installed workload estate, accumulated database and industry-model intellectual property, and expensed software research; their value should be inferred from retention, pricing, gross profit, and cash flow—not from development cost alone. The mirror-image risk is technical debt. Legacy code, customer-specific complexity, old deployment patterns, and contractual maintenance obligations can be economic liabilities disguised as installed-base assets.

Security and tax status

The security is ordinary common stock of a Delaware corporation listed on the NYSE; it is not an ADR, MLP, partnership, or K-1 issuer. Teradata has not paid a regular dividend and returns capital primarily through share repurchases. Conventional U.S. corporate tax reporting applies to shareholders when distributions or sales create taxable events. [S1] [S12]

Verdict: Teradata has a comprehensible and predominantly recurring business with real installed-base assets. Its revenue quality is below pure ratable SaaS because upfront licenses, consumption, hardware rentals, and loss-making consulting complicate timing, margin, and capital intensity.

Industry Dynamics

Teradata participates in overlapping markets for analytical databases, enterprise data warehouses, lakehouses, data governance, public-cloud data infrastructure, and AI-enabling data platforms. A single market-size estimate would create false precision because these categories overlap with hyperscale infrastructure and enterprise applications. The economically relevant observation is that customers continue to spend heavily on modern data platforms while deciding which vendor controls storage, compute, governance, developer tooling, and AI context. [S1] [S13] [S14]

Market growth and geography

Market growth is global rather than primarily domestic: Teradata generated approximately half of 2025 revenue internationally, while Snowflake and MongoDB continued to report rapid growth in their latest July quarters. Snowflake’s product revenue was approximately $1.492 billion, up 37%, and MongoDB’s total revenue was approximately $772 million, up 30%. Those figures do not define the total addressable market, but they demonstrate that enterprise data spending is growing elsewhere even while Teradata’s total revenue is stagnant or declining. [S1] [S13] [S14]

Demand varies by geography and workload. Data-sovereignty requirements, public-cloud availability, latency, national-security rules, and regulation can favor on-premises or sovereign deployments. Currency can materially change reported ARR because about half of Teradata’s revenue is international. The company’s 2025 Total ARR growth was 3% reported but only 1% in constant currency; cloud ARR growth was 15% reported and 13% in constant currency. Investors should therefore compare reported and constant-currency performance rather than credit translation as organic demand. [S1]

Competitive structure

The industry has several classes of competitor. Hyperscalers—AWS, Microsoft Azure, and Google Cloud—bundle storage, compute, databases, governance, AI services, and distribution. Snowflake and Databricks compete for cloud data and AI workloads with architectures that separate storage and compute and increasingly support open formats. Oracle and other legacy database suppliers defend integrated installed bases. MongoDB is more focused on developer-oriented operational data, making it an adjacent rather than direct warehouse peer. Teradata itself names AWS, Databricks, Google Cloud, Microsoft Azure, Snowflake, and legacy suppliers as competitors. [S1]

Industry profitability is bifurcated: hyperscalers and scaled platforms can earn attractive recurring economics, but specialists must fund substantial research, sales, support, and cloud commitments. Snowflake’s latest product gross margin was approximately 71%, but it remained GAAP loss-making and incurred more than $400 million of quarterly stock compensation. MongoDB’s subscription gross margin was about 77%, but its stock compensation was also substantial. Teradata is GAAP profitable and trades at a much lower valuation, yet its 68% 2025 recurring gross margin and low growth prevent the conclusion that its economics are superior. [S1] [S13] [S14]

Peer comparisons must therefore separate growth, retention, gross margin, stock compensation, and capital intensity. Snowflake’s 126% net revenue retention and MongoDB’s 122% net ARR expansion show far stronger expansion than Teradata’s 108% 2025 cloud net expansion. Conversely, those peers use more equity compensation and serve somewhat different workloads. They are evidence about where incremental spending is going, not interchangeable valuation comparables. [S1] [S13] [S14]

Barriers to entry and switching costs

Building an enterprise-grade distributed database requires query optimization, workload management, governance, security, uptime, connectors, observability, support, and years of reference customers. Those engineering and trust barriers are real. For mission-critical workloads, customers must also validate reconciliation, performance, regulatory controls, and disaster recovery. A new vendor cannot reproduce those credentials merely by matching a feature checklist.

The barriers protect incumbent workloads more than they protect new workload creation. Hyperscalers already possess infrastructure, customer relationships, identity systems, billing, and developer ecosystems. Open-source software lowers feature-development costs. Systems integrators are paid to help customers migrate. Open table formats can reduce data-movement friction. Capital is therefore abundant even though technical entry is difficult.

Industry profitability is divided between vendors with ecosystem control and specialists facing heavy research and selling requirements; at least six major competitive platforms matter, and barriers protect embedded production workloads more effectively than greenfield demand. [S1] [S13] [S14]

Supply-side capital cycle

This is not a physical capital cycle with easily observed factories. Supply expands through R&D, cloud-capacity commitments, AI inference costs, stock compensation, acquisitions, and bundled pricing. Hyperscalers can subsidize database tools to drive infrastructure consumption. Venture-backed and public software vendors can invest aggressively despite GAAP losses. Open-source projects can replicate capabilities without a conventional corporate cost of capital.

The result is a market where end-demand growth does not guarantee attractive returns for every vendor. Customers gain more credible substitutes, procurement leverage, and architectural flexibility. Snowflake disclosed rising infrastructure and AI inference costs, while MongoDB carries cloud-capacity commitments. Teradata finances some customer equipment. The relevant capital-cycle question is not whether data volumes grow—they almost certainly do—but whether differentiated software value grows faster than the infrastructure and distribution cost required to monetize it. [S1] [S13] [S14]

Open formats and ecosystem convergence

Competition is becoming more intense because open formats, hyperscaler bundling, lakehouse convergence, and AI-platform spending expand the number of credible substitutes even as demand grows. Teradata’s OneLake integration uses Iceberg APIs to query data without mandatory duplication or migration. That is strategically adaptive: it may preserve Teradata compute in Microsoft-centered estates. It is also disconfirming evidence for a strong proprietary-data moat because customers can keep data elsewhere and compare processing engines more easily. [S8]

Interoperability can therefore be both offensive and defensive. It reduces the friction of using Teradata on external data, potentially expanding addressable workloads. It also reduces the cost of replacing Teradata for selected workloads. The economic outcome must be measured through consumption, attach rates, renewal, and gross margin rather than connector announcements.

Regulation and sovereignty

Privacy, cybersecurity, AI governance, export controls, sanctions, and data-residency requirements raise compliance costs. They may favor Teradata’s hybrid and Factory positioning when customers cannot place sensitive workloads in a public cloud. That opportunity is not exclusive. Oracle, Microsoft, IBM, hyperscaler appliances, and open-source private-cloud stacks can offer residency and governance alternatives. Regulation raises the value of trusted deployment but does not guarantee Teradata wins.

Foreign low-cost threat

Foreign low-cost labor cannot directly replicate Teradata’s installed enterprise estate, but global engineering supply and offshore services can reduce competitors’ development costs and compress consulting prices. Teradata outsources hardware assembly to Flex and operates with a global workforce. The more material threat than wage arbitrage is bundling by vendors that already monetize cloud infrastructure. Hardware supply is a separate issue: AI-related demand for memory and other components may pressure availability or pricing for Factory and rental equipment. [S1] [S2]

Verdict: The market is large and growing, but competitive supply and ecosystem power are growing faster than Teradata’s revenue. Industry structure favors hyperscalers and scaled platforms; Teradata’s opportunity is a defensible hybrid niche, not broad control of the data stack.

Competitive Position

Teradata’s competitive position rests on high-scale workload performance, mixed-workload management, industry and data-model expertise, enterprise trust, and migration friction. Its weakness is distribution into greenfield cloud-native workloads. A customer can continue to rely on Teradata for governed production analytics while allocating incremental lakehouse, application, and AI spending elsewhere.

Performance and workload management

The historical differentiation is the ability to execute complex analytical workloads across large data estates with high concurrency. The Autonomous Knowledge Platform adds always-on Active Compute and on-demand Elastic Compute, with the stated goal of separating stable production workloads from bursty experimentation. The architecture is plausible, but the reviewed evidence contains no independent current benchmark reconciling price, workload complexity, infrastructure cost, service levels, and migration effort across Teradata, Snowflake, Databricks, Oracle, or hyperscaler offerings. [S7]

Performance claims should therefore be treated as management hypotheses. Decision-useful evidence would include independent workload benchmarks, customer cost comparisons, renewal economics, and win-loss data. The financial proxy is whether customers expand ARR while recurring gross margin holds. Q2’s 1% Total ARR growth and flat sequential cloud ARR do not demonstrate broad performance-led share gains. [S2]

Hybrid deployment

Management said more than half of public-cloud customers also use Teradata in hybrid configurations. Factory extends the new platform onto Dell and NVIDIA infrastructure for on-premises and sovereign environments. That gives Teradata a coherent answer for customers that cannot migrate every workload to a public cloud. It also reuses the company’s installed-base strength rather than requiring it to win solely as a cloud-native vendor. [S6] [S19]

Hybrid differentiation should not be overstated. Oracle, Microsoft, IBM, hyperscaler appliances, and open-source stacks offer hybrid approaches. Factory also reintroduces supply, inventory, depreciation, and lease-funding considerations. A dollar of Factory ARR may not have the margin or capital intensity of a dollar of public-cloud software. The product becomes a moat only if it improves Total ARR retention or expansion at attractive returns.

Switching costs

Customer switching costs are high for deeply integrated production workloads but lower for new workloads; therefore they support retention more reliably than expansion or new-logo growth. Schemas, stored procedures, data pipelines, governance rules, dashboards, access controls, workload policies, and employee skills must be translated and validated. Parallel operation may be required, and migration creates operational and regulatory risk. [S1] [S2]

Those costs are a genuine narrow moat. They give Teradata time to improve products and make replacement decisions expensive. They do not prevent gradual erosion. Automation reduces conversion effort, integrators benefit from migration projects, and open formats reduce data relocation. If switching costs created broad pricing power, total revenue and recurring gross profit would be expanding rather than declining.

The financial evidence supports a moat around the core and weakness at the perimeter. Recurring revenue remained almost 87% of sales and Total ARR returned to modest growth in 2025. Against that, revenue declined 13% from 2021 through 2025, cloud net expansion fell from 117% to 108%, and cloud-native competitors grew much faster. [S1] [S13] [S14] [S18]

Context, governance, and AI

Teradata argues that governed business context, semantics, lineage, and industry models can make AI outputs and agents more reliable. That is a reasonable enterprise requirement. Decades of work in banking, telecommunications, retail, transport, healthcare, and government may give the company knowledge that is not captured in its balance sheet.

The missing link is monetization. AI functionality may use capacity already purchased, improve retention without separate pricing, or require services before recurring consumption appears. Management said the new products were not material to first-half 2026 results and warned that ARR can lag initial customer use. Engagement counts and product launches should therefore update relevance before they update revenue or ROIC. [S5] [S6] [S7]

Brand and trust

The brand is economically relevant for enterprise trust and installed-base retention, but there is no evidence of broad pricing power: revenue fell 13% from 2021 through 2025 despite continued product investment. In mission-critical data procurement, familiarity can shorten security review, reassure decision-makers, and improve access to senior buyers. It is not a consumer brand that automatically raises willingness to pay. [S1] [S18]

Brand value must manifest as retention, price realization, lower selling cost, or superior margins. Teradata’s recurring base shows continuing trust, while the growth record shows that trust has not translated into leadership for new workloads.

Interoperability and OneLake

The Microsoft OneLake integration allows Teradata to query OneLake tables through Iceberg-compatible interfaces without requiring data duplication. Strategically, this converts an all-or-nothing platform choice into a workload decision. Teradata can remain the analytical engine for selected high-value workloads even if storage and surrounding tooling are Microsoft-controlled. [S8]

The same architecture weakens proprietary lock-in. If data is portable and multiple engines can access it, the customer can shift compute more readily. OneLake therefore reduces migration risk in both directions. The deciding evidence will be incremental consumption and renewals, not technical availability.

Go-to-market and consulting

Teradata increasingly relies on cloud providers, systems integrators, and technology partners. A partner ecosystem can expand reach and reduce the cost of internal consulting. Yet consulting revenue declined 19% in 2025 and another 24% in Q2 2026, while Q2 GAAP consulting gross profit was negative. That may be an intentional transition, but it may also indicate reduced influence over implementation choices. [S1] [S2]

The company does not disclose partner-sourced ARR, implementation-cycle improvements, or competitive win rates. Without those metrics, partnership claims remain strategically plausible but economically unverified.

Competitive direction by workload

For existing large production warehouses, Teradata remains credible and sticky. For migration of its own base to public cloud, it has achieved substantial cloud ARR but slowing expansion. For greenfield workloads, peer growth suggests it is not a leading share winner. For sovereign and hybrid AI, Factory offers a plausible niche with as-yet-undisclosed economics.

The nature of competition is workload-by-workload rather than winner-take-all: customers can retain Teradata for governed production analytics while placing new applications, lakehouse data, and AI experimentation on other platforms. Coexistence can stabilize the installed base while capping wallet-share growth. [S1] [S8]

Verdict: Teradata has a narrow installed-base moat grounded in switching costs and trusted workload performance. It currently preserves value better than it creates growth, and evidence does not support a broad innovation or distribution moat.

Growth History and Forward Opportunities

Revenue declined from $1.917 billion in 2021 to $1.795 billion in 2022, recovered to $1.833 billion in 2023, then fell to $1.750 billion in 2024 and $1.663 billion in 2025. The four-year compound decline was approximately 3.5%. Gross profit fell from $1.186 billion to $987 million. Operating income was more resilient—$231 million, $118 million, $186 million, $209 million, and $205 million—because cost reductions offset revenue pressure. [S1] [S12] [S16] [S17] [S18]

The mix transition is more favorable than the total revenue trend but less favorable than cloud ARR alone suggests. At year-end 2025, Public Cloud ARR was $701 million, up from $609 million; subscription ARR was $735 million, down from $766 million; and maintenance ARR was $86 million, down from $99 million. Total ARR rose to $1.522 billion from $1.474 billion, but constant-currency growth was only 1%. Cloud growth therefore contains migrations from other Teradata categories as well as incremental spending. [S1]

At June 2026, Public Cloud ARR was $686 million, subscription ARR $750 million, maintenance ARR $73 million, and Total ARR $1.509 billion. Compared with Q1, cloud ARR was unchanged while Total ARR increased $17 million. Year-over-year performance remained positive, but the absence of sequential cloud growth is inconsistent with a strong current inflection. [S2] [S21]

Management guides 2026 Total ARR growth of 2%-4%. Applied to the $1.522 billion 2025 base, the midpoint implies about $1.568 billion at year-end, requiring roughly $59 million of growth from Q2. The range requires $43-$74 million. Management expects modest sequential growth in Q3 and most of the increase in Q4, when significant renewals occur. This may be achievable, but the concentration raises execution risk. [S3] [S5]

The product outlook is promising in breadth but unproven economically: public cloud, Factory, AI Studio, Tera, agents, elastic compute, and OneLake interoperability can improve relevance, yet management said they did not materially affect first-half 2026 results and disclosed no attributable ARR. [S5] [S7] [S8] [S19]

Installed-base expansion

The lowest-risk growth route is selling additional workloads to existing customers. Teradata has already passed procurement, governance, and security review in those accounts. AI Studio, elastic compute, and OneLake access may generate incremental usage without requiring a wholesale platform decision. Existing customer familiarity can also reduce implementation cost.

The disconfirming possibility is cannibalization. A customer may move an existing workload from subscription ARR to public-cloud ARR without increasing total spending. New AI features may consume unused contracted capacity. Growth should therefore be measured through Total ARR, recurring gross profit, and cohort expansion—not cloud reclassification.

Retention improvement

Management attributes recent Total ARR stabilization partly to better retention. Year-end 2025 Total ARR growth and Q2 2026 positive growth are consistent with that claim. However, Teradata does not disclose total-company gross or net retention. The annual cloud net expansion rate fell from 117% in 2024 to 108% in 2025. Improved retention may be real while expansion weakens; the public data cannot separate them. [S1] [S5]

This is the single most important disclosure gap. Gross retention would show whether switching costs are holding. Net retention would show whether surviving customers expand. Migration decomposition would show whether cloud growth is incremental or merely deployment movement.

Hybrid and sovereign AI

Factory targets customers that require local control, low latency, sovereignty, or integration with existing on-premises estates. Dell and NVIDIA infrastructure can make the offering commercially credible and may defend accounts against cloud-only competitors. Demand could be meaningful in financial services, government, healthcare, telecommunications, and other regulated sectors. [S19]

The economics are unknown. Factory requires equipment, supply commitments, installation, support, and potentially finance leases. Management indicated that product supply and pricing could matter more in 2027 than in first-half 2026. Investors should not apply cloud-software margins until equipment returns and contract structure are disclosed. [S2] [S5]

Open ecosystem participation

OneLake and Iceberg integration may allow Teradata to monetize analytical compute even when customers store data outside Teradata. This could reduce the risk that a storage decision forces complete platform displacement. It also broadens the addressable installed base to Microsoft-centered customers. [S8]

The counterargument is that open access commoditizes engines and weakens switching costs. A connector is table stakes if competing engines offer the same interoperability. Success requires incremental consumption and durable renewal economics.

Cloud-margin improvement

Q2 2026 GAAP recurring gross margin improved 50 basis points year over year, and management cited better public-cloud unit margin. If hosting utilization, contract pricing, and infrastructure purchasing improve, cloud growth need not cause perpetual margin decline. [S2] [S3]

Mix remains a headwind. Cloud is still less profitable than upfront software, while public-cloud ARR growth has slowed. The company must improve unit margins without underinvesting in service quality or using longer commitments that create hidden capacity risk.

Per-share growth

Low-single-digit company growth can produce faster per-share growth if margins hold and repurchases exceed issuance. This is financial compounding, not competitive share gain. It works only when shares are repurchased below intrinsic value and the company does not consume excess cash through acquisitions or offsetting equity grants.

The 2025 share reduction demonstrates that the mechanism can work, but the high burn rate and stock compensation show its cost. Per-share growth should be evaluated after awards and lease investment.

Forecast credibility

Management’s 2024 miss is the strongest reason to haircut the current plan. Initial constant-currency Total ARR guidance of 4%-8% became an actual decline of 4%, a roughly ten-point shortfall at the midpoint. The company responded with restructuring and a deployment-neutral strategy, so the current organization is not identical. Nonetheless, back-end-loaded guidance deserves less weight until signed renewals and realized ARR are visible. [S12] [S22] [S23]

Verdict: The realistic growth opportunity is retention plus installed-base expansion. Product breadth may stabilize the franchise, but current evidence does not support sustained mid-single-digit growth or AI-led acceleration.

Financial Quality

Five-year operating record

The series was reconciled between Company Financials and annual filings. Revenue and gross profit declined, while operating income recovered through cost control. SG&A fell from $635 million in 2023 to $502 million in 2025, whereas R&D declined more modestly from $294 million to $280 million. The company protected product investment better than selling and administrative spending, but the return on that investment is not yet visible in revenue. [S1] [S12] [S16] [S17] [S18]

Earnings are not at a conventional cyclical high or low; they are near a self-help margin high within a declining revenue base, making further cost-led gains harder unless ARR growth resumes. External demand for data platforms remains strong, so Teradata’s earnings cycle is predominantly company-specific transformation rather than an industry recession. [S1] [S13] [S14]

Current performance

Q2 2026 revenue was $410 million: $363 million recurring, $39 million consulting, and $8 million perpetual licenses, hardware, and other. Gross profit was $243 million, operating income $48 million, and net income $46 million. GAAP operating margin reached 11.7%, versus 5.9% a year earlier. Non-GAAP operating margin was 21.5%, largely because management excluded $33 million of stock compensation and $7 million of reorganization and other costs. [S2] [S3]

The non-GAAP measure is useful for comparing expense control but is not owner profit. Stock compensation is recurring, and restructuring has recurred during the transformation. A valuation based on adjusted EPS should separately account for dilution and the cash required to offset it.

SAP settlement normalization

First-half 2026 GAAP net income of $381 million and cash from operations of $507 million are not representative. The SAP settlement generated $480 million of other income and $121 million of legal and contingent fees, producing a $359 million pretax net benefit. Management’s full-year cash-flow guidance includes an estimated $315 million after-tax net benefit. [S2] [S3]

Adjusted free cash flow removes the $480 million gross receipt and adds back associated legal fees and settlement-specific taxes. On that basis, first-half adjusted free cash flow was $158 million rather than reported free cash flow of $495 million. Full-year GAAP EPS guidance of $4.43-$4.51 is dominated by the legal recovery; adjusted EPS guidance of $2.65-$2.73 is a better operating comparison, subject to recurring exclusions. [S2] [S3]

The settlement is economically valuable because it increased cash and enabled debt repayment. It is not evidence of customer demand or product competitiveness. Balance-sheet value and operating performance must remain separate.

Cash conversion and income divergence

Net income and operating cash flow diverged materially: in 2025 net income was $130 million versus $305 million of operating cash flow, aided by $90 million of depreciation and amortization, $112 million of stock compensation, deferred-revenue movements, and other working-capital changes. In first-half 2026, settlement accounting dominated the divergence. [S1] [S2]

The 2025 divergence is economically mixed. Deferred revenue is favorable financing but carries future obligations. Depreciation partly reflects customer equipment that must be replaced. Stock compensation is noncash in the cash-flow statement but transfers ownership or requires repurchases. Cash generation is stronger than GAAP earnings, but not every add-back is distributable.

Capital intensity

Reported cash capital expenditure was only about $20 million in 2025, suggesting an asset-light model. That is incomplete. Teradata paid $69 million of finance-lease principal and used leased equipment to support customer services. Rental revenue, including hardware maintenance, was $212 million. [S1]

Reported capex makes the business look asset-light, but customer-equipment finance leases are economically capital investment; after $69 million of 2025 lease principal, a conservative owner-earnings proxy was about $216 million rather than the stated $285 million of free cash flow. This is not a definitive maintenance-capex estimate because principal includes financing timing and growth assets. A reasonable valuation should test both reported FCF and a lease-adjusted range.

The physical capital intensity remains modest relative to manufacturing, but materially higher than that of a pure software subscription. Factory may raise equipment intensity if it becomes successful.

Profitability and ROIC

GAAP operating margin recovered from 6.6% in 2022 to 12.3% in 2025. Company Financials’ standardized ROIC increased from 5.9% in 2022 to 18.8% in 2025. Conventional balance-sheet ROIC can look even higher because deferred revenue finances operations, repurchases reduce book equity, and internally developed software is not capitalized. [S1] [S12]

A research-adjusted calculation should capitalize a portion of historical R&D. Under a five-year straight-line life and half-year convention, the year-end 2025 research asset is roughly $720 million and 2025 amortization roughly $270 million. Adjusted EBIT would be near $215 million after adding back $280 million of R&D and deducting amortization. Adding the research asset and finance leases to invested capital produces a rough research-adjusted ROIC range of approximately 12%-16%, depending on tax rate, working-capital treatment, useful life, and impairment assumptions. This is an analyst estimate, not a reported metric. [S1] [S12] [S16] [S17] [S18]

Business profitability is positive, but conventional ROIC is flattered by expensed R&D, deferred-revenue financing, and a repurchase-compressed equity base; a consistent research-adjusted estimate is plausibly in the low-to-mid teens rather than a superficial 50%-plus result.

The adjusted range may still exceed the cost of capital, but the margin is not wide enough to establish a strong growth moat. A longer useful life would raise the research asset and reduce annual amortization; a shorter life or impairment would reduce capital but also call past research value into question.

Accounting quality

Revenue recognition requires judgment because contracts can contain software, cloud services, support, hardware, and consulting. The auditor identified revenue recognition for certain arrangements as a critical audit matter, including identification of performance obligations and standalone selling prices. Upfront recognition of term licenses is permitted under GAAP but creates more volatility than ratable SaaS. [S1]

Accounting is mixed rather than uniformly conservative or aggressive. Immediate R&D expensing depresses current earnings and understates invested capital. Upfront term-license recognition accelerates revenue relative to service duration. Stock compensation is expensed under GAAP but removed from management’s key adjusted measures. Finance-lease equipment is recognized on the balance sheet, yet principal falls outside management-defined FCF.

There were no significant changes in critical accounting policies in the first half of 2026. The larger comparability issue is transaction mix and settlement accounting, not a new accounting rule. The SEC’s 2022 comment letter asked for more specific disclosure of cloud net expansion and monitoring of other recurring revenue. Teradata now reports annual cloud expansion, but still does not disclose comprehensive total-company retention. [S2] [S20]

Balance sheet, liquidity, and obligations

At June 2026, cash was $414 million, the term loan had been repaid, no revolver borrowing was outstanding, and $400 million of revolving capacity was available. Finance-lease liabilities were $91 million. Net cash after finance leases was therefore approximately $323 million. [S2]

The balance sheet also contained $111 million of pension and postemployment liabilities. Year-end 2025 purchase obligations were $553 million, including cloud hosting, IT, and other commitments. Those purchase obligations are not conventional funded debt, but they represent future cash claims. Customer indemnities, operating leases, and required equipment replacement add further economic obligations. The company disclosed no special-purpose financing arrangements. [S1] [S2]

The balance sheet is strong enough that ordinary operating disappointment is unlikely to create near-term solvency risk. The central question is how management deploys the repaired liquidity.

Verdict: Financial quality is adequate and improving, not exceptional. Recurring cash generation and net cash are strengths; declining gross profit, recurring stock compensation, lease-funded equipment, timing-sensitive revenue, and uncertain research returns qualify the headline yield and ROIC.

Capital Allocation

Teradata’s recent allocation record consists of R&D, restructuring, customer-equipment financing, repurchases, and debt reduction. There is no regular dividend and no material recent acquisition program. The SAP settlement gave management a one-time opportunity to remove leverage without consuming ordinary operating cash flow.

Reinvestment

R&D was $294 million in 2023, $284 million in 2024, and $280 million in 2025, equal to 16.8% of 2025 revenue. Q2 2026 R&D increased to $75 million from $71 million as the company invested in cloud and AI. This is substantial spending for a mature vendor. [S1] [S2]

Revenue declined while R&D remained high, so expenditure alone cannot be treated as productive reinvestment. The relevant outputs are retention, incremental ARR, gross-margin stability, and workload wins. Factory also requires physical equipment and supply-chain support beyond software research.

Free cash flow and use

Free cash flow was $355 million in 2023, $277 million in 2024, and $285 million in 2025. After finance-lease principal, conservative proxies were approximately $273 million, $206 million, and $216 million. Management plans to use about 50% of adjusted free cash flow, excluding the settlement, for repurchases. [S1] [S5]

Free cash flow is substantial but overstates distributable owner earnings if customer-equipment lease principal is ignored; management has used much of residual cash for repurchases and used the 2026 legal recovery to eliminate term debt. [S1] [S2] [S5]

The allocation philosophy is therefore shareholder-return oriented, but not purely distributive. R&D is already expensed within operating cash flow, while equipment leases, restructuring, and working capital absorb additional resources.

Buybacks and dilution

Repurchases were $308 million in 2023, $215 million in 2024, and $140 million in 2025. Teradata repurchased approximately 5.8 million shares at an average $36.99 in 2024 and 5.8 million at $24.34 in 2025. During first-half 2026 it repurchased 2.5 million shares at $29.76 for $74 million, leaving $430 million of authorization. [S1] [S2]

The record is mixed. The 2024 purchases are materially above the current price, the 2025 purchases were below it, and the 2026 purchases were close to it. This indicates a program that has reduced shares but has not consistently demonstrated price discipline.

Share repurchases have produced a real net reduction despite equity issuance: diluted weighted-average shares fell from 112.9 million in 2021 to 96.6 million in 2025, and issued shares were 93.4 million at June 2026; however, $112 million of 2025 stock compensation and a 6.24% equity-plan burn rate consumed a meaningful portion of the program. [S1] [S2] [S4]

The 2025 year-end share count declined by only about 2.6 million despite roughly 5.8 million repurchased shares, illustrating that awards and related issuance offset more than three million gross repurchases. The reduction still created per-share value if the shares were bought below intrinsic value, but gross repurchase dollars should not be presented as equivalent to shareholder yield.

Insider transactions and issuance

The reviewed Forms 3, 4, and 5 primarily show grants, vesting, withholding, and sales rather than open-market executive buying. Director Michael Gianoni sold 25,000 shares at $29.965 in February 2026. August director filings reported $0 restricted-unit awards under the director compensation program, not purchases. The Q2 filing separately disclosed 348,235 shares transferred by Section 16 officers for tax withholding at an average $33.99; those were not discretionary company repurchases or open-market purchases. [S2] [S15] [S26]

A review of relevant insider filings found no decision-useful pattern of open-market executive purchases; equity issuance is material through restricted awards, while many reported acquisitions are compensation grants rather than insider conviction signals.

No single sale proves adverse information, but the absence of meaningful open-market buying limits the bull case based on insider alignment.

Compensation policy

The 2025 annual incentive weighted constant-currency Total ARR growth at 40%, Public Cloud ARR growth at 20%, and non-GAAP operating-margin dollars at 40%. Actual Total ARR growth of 0.74% earned 87% on its component; cloud growth of 13.1% was below threshold and earned zero; non-GAAP operating-margin dollars of $342 million earned 162%. The combined payout was 100%. [S4]

Compensation policy ties annual incentives to Total ARR, Public Cloud ARR, and non-GAAP operating margin, while long-term awards use Total ARR, Rule of 40, and free cash flow; the design is directionally aligned, but a key margin measure excludes recurring stock compensation. [S4]

Target calibration is a larger concern. The $322 million 2025 target for non-GAAP operating-margin dollars was below the preceding year’s result, making above-target payout easier during a cost program. The committee also reset goals for the 2023-2025 and 2024-2026 long-term plans after 2024 business challenges; the adjusted 2023-2025 plan paid 74%. Maximum payout was reduced, which partially mitigates the reset, but changing goals after underperformance weakens the original performance contract. [S4]

The 2025 equity burn rate rose to 6.24% from 3.29% in 2024, and fully diluted plan overhang was 11.23% in March 2026. Those figures make net dilution a key monitoring item.

Management behavior suggests genuine balance-sheet discipline alongside incentive preservation: debt was eliminated and shares declined, but earlier repurchases were price-insensitive and long-term compensation targets were reset after underperformance. [S2] [S4]

Governance and Lynrock

Lynrock Lake held approximately 9.9% and entered a cooperation agreement that provided for board refreshment. Melissa Fisher joined the board, and Bernd Leukert was appointed effective August 1, 2026, expanding the board from nine to ten members. Leukert brings software, cloud, AI, and financial-services experience. [S4] [S11] [S27]

The agreement creates accountability and relevant expertise, but calling it unrestricted activist pressure would overstate the facts. Lynrock accepted standstill and voting provisions. Governance influence is real; continuing public pressure is constrained during the agreement.

Acquisitions and dividends

No material acquisition was disclosed during 2023-2025, and goodwill remained near $400 million. There is therefore no recent acquisition cohort whose return can be credibly calculated. Historical goodwill indicates earlier purchased assets, but current filings do not provide enough asset-level economics to isolate their return. [S1] [S12]

Teradata pays no regular dividend and has stated that it does not expect to initiate one in the immediate future; dividend coverage is therefore not applicable, and repurchases remain the regular direct capital-return mechanism. [S12]

The repaired balance sheet creates optionality but also raises risk. Management mentioned opportunistic strategic M&A on the Q2 call. A large acquisition would require a new underwriting because the current valuation thesis relies partly on net cash and low financial risk. [S5]

Verdict: Allocation has improved through debt elimination and net share retirement. High equity burn, compensation resets, uneven repurchase timing, and renewed M&A optionality prevent a clean governance endorsement.

Changes and Headwinds — Last Two Years

The operating environment and Teradata’s response changed materially from 2024 through 2026. Customers reassessed data architectures as generative AI increased the importance of governance, context, model access, and open data. Management described longer decisions and staged purchases. At the same time, cloud-native competitors continued to expand and open table formats reduced proprietary isolation. [S1] [S4] [S8] [S13] [S14]

The business environment changed materially: buying shifted toward hybrid deployment and AI architecture reviews, open formats reduced proprietary isolation, and hyperscalers and lakehouse vendors intensified competition. [S1] [S4] [S8]

2024 execution break

The most important change was the failure of the 2024 plan. Management initially expected constant-currency Total ARR growth of 4%-8% and Public Cloud ARR growth of 35%-41%. By August, Total ARR guidance had been reduced to a decline of 2%-4%, and actual full-year Total ARR declined 4% in constant currency while cloud ARR grew 18%. [S12] [S22] [S23]

This was not a minor timing miss. It indicated that migrations, retention, expansion, or sales execution were materially weaker than planned. Management attributed part of the shortfall to elongated cycles and customers reconsidering architecture amid AI. The magnitude and peer growth suggest internal execution also mattered.

Restructuring and margin repair

The 2024 restructuring targeted approximately 9%-10% of employees, $75-$80 million of annual operating-expense savings, and $45-$50 million of severance cash costs. Headcount fell from about 7,200 in 2021 to 5,100 at year-end 2025. SG&A declined from $635 million in 2023 to $502 million in 2025. [S1] [S18] [S22]

Results over the last two years were driven by both external architecture changes and internal action, but margin improvement was chiefly internal cost reduction while weak relative growth reflected execution as well as customer caution. [S2] [S4] [S13] [S14]

Q2 2026’s 11.7% GAAP margin compared with 5.9% a year earlier confirms that cost actions affected earnings. The question is whether the company can reinvest enough to protect products while preserving those margins.

Strategic pivot

Teradata shifted from emphasizing public-cloud migration to a deployment-neutral Total ARR strategy. The proxy increased the incentive weight on Total ARR and reduced emphasis on Public Cloud ARR so management would not favor one deployment. The Autonomous Knowledge Platform, Factory, AI Studio, and OneLake integration embody that approach. [S4] [S7] [S8] [S19]

The pivot may fit customer demand better and protect profitable on-premises workloads. It also implicitly acknowledges that public-cloud migration alone did not generate sufficient total-company growth. Total ARR is the right top-line metric only if the company also discloses retention, mix, and margins.

Leadership and board changes

Claire Bramley left the CFO role and John Ederer became CFO. Sumeet Arora became chief product officer, and Josh Fecteau became chief data, analytics, and information officer. The Lynrock agreement produced board refreshment, including Melissa Fisher and Bernd Leukert. [S4] [S11] [S27]

Important market, facility, and management changes include the hybrid strategic pivot, Factory’s Dell/NVIDIA architecture, a smaller workforce, new financial and product leadership, and board refreshment under the Lynrock agreement. [S4] [S11] [S19] [S27]

These changes may improve accountability and product focus, but they also create execution risk during a critical renewal period.

Litigation and balance-sheet repair

The SAP dispute ended in February 2026. Teradata received $480 million, incurred $121 million of related fees, and used cash to repay the $450 million term loan. This removed litigation and refinancing risk but did not improve recurring demand. [S2]

The 2025 10-K described a securities class action concerning statements about 2023 ARR expectations, while the Q2 2026 filing stated that pending or threatened litigation was not expected to have a material adverse effect. A complaint is an allegation rather than an adjudicated finding. [S1] [S2]

Accounting and current headwinds

Accounting policies did not materially change in first-half 2026; the larger comparability issue is transaction mix and timing, particularly upfront on-premises license recognition and settlement accounting. [S2]

Current headwinds include Q3 revenue declines, Q4-weighted ARR, flat sequential cloud ARR, shrinking maintenance, low consulting profitability, public-cloud mix pressure, stock compensation, and equipment-component risk. Offsets include improved reported retention, lower SG&A, better cloud unit margin, net cash, and open-platform integration.

Verdict: Teradata responded materially to a changed market, but verified improvement is mainly cost, governance, and balance-sheet driven. Product strategy has not yet generated measurable incremental growth.

Risk Analysis

Factors that could cause the stock to decline include an ARR miss, continued cloud deceleration, weak Q4 conversion, falling recurring margin, unmonetized AI products, dilution, or a value-destructive acquisition. A broad software multiple contraction would compound those company risks. [S1] [S2] [S3] [S4]

Gradual erosion

The dominant downside is not immediate failure but a melting installed base. Cloud migrations could approach exhaustion while maintenance and subscription attrition continue. Cost cuts could preserve profit temporarily, making cash flow look resilient until recurring gross profit falls below the fixed R&D and support base. This pathway deserves more weight than a conventional recession scenario because peers show continuing data-platform demand.

Forecasting and revenue timing

Upfront term-license recognition and cloud consumption make quarterly forecasting difficult. They also create room for repeated timing explanations. The 2024 guidance miss demonstrates that forecast risk is empirical, not theoretical. A Q3 revenue decline may be accounting timing; a Q3 or Q4 ARR miss would be more serious. [S5] [S22] [S23]

Cash-quality risk

Management-defined FCF can be achieved while lease principal, equity awards, or deferred-revenue reversals consume economic value. The safeguard is a bridge from CFO to cash capex, lease principal, net dilution, and working capital. If rental ARR grows without a corresponding return on leased assets, reported FCF will overstate the business’s distributable capacity.

Catastrophic-loss pathway

A catastrophic loss would require several failures together—accelerated core-workload displacement, a major security or reliability event, and poor deployment of cash—because current net cash and recurring revenue make a single weak quarter insufficient. [S1] [S2]

A severe data-integrity event could be unusually damaging because customers rely on Teradata for governed and operational decisions. Claims, remediation, loss of trust, and accelerated migration could arrive together. If management simultaneously re-levered for an acquisition, the balance-sheet buffer would disappear.

Total-loss pathway

The chance of total loss is low because Teradata is net-cash and cash-generative; the plausible route is not ordinary competition alone but competition combined with fraud, operational catastrophe, extreme legal liabilities, or renewed leverage. [S1] [S2]

At June 2026, $414 million of cash, no term debt, an undrawn $400 million revolver, and a $1.5 billion recurring-revenue base make literal insolvency remote. Finance leases, pensions, and purchase commitments are meaningful but not currently sufficient to erase the enterprise alone.

Risk mitigants

Net cash, recurring revenue, no customer above 10%, geographic diversity, cost flexibility, and an undrawn revolver reduce financial distress risk. Switching costs slow displacement. These mitigants protect survival, not necessarily intrinsic value. A profitable shrinking company can still produce poor returns if the multiple and cash flows decline together.

Verdict: The principal risk is gradual competitive erosion hidden by self-help cash flow. Solvency risk is low today, but it would rise materially if management used the repaired balance sheet for a large acquisition.

Valuation Discussion

The valuation date is September 11, 2026, the latest verified market close before publication. At $27.84 and 93.4 million issued shares, equity value was approximately $2.60 billion. Adding $91 million of finance leases and subtracting $414 million of cash produces enterprise value near $2.28 billion. Pension and purchase obligations are discussed as claims but not included in that standard bridge. [S2] [S9]

Current multiples and embedded expectations

Management guides 2026 Total ARR growth of 2%-4%, recurring-revenue growth of flat to 2%, total-revenue change of -2% to flat, adjusted EPS of $2.65-$2.73, and adjusted free cash flow of $330-$350 million. The current price implies approximately 10.3 times midpoint adjusted EPS, 7.6 times midpoint adjusted FCF, and enterprise value near 1.38 times the midpoint revenue implied by guidance. [S3]

Those are not directly comparable economic measures. Adjusted EPS excludes stock compensation and restructuring. Adjusted FCF includes the operating benefit of noncash stock compensation and excludes finance-lease principal. A lease-adjusted owner-earnings range of $270-$290 million implies 9.0-9.6 times equity value before crediting excess cash, or a 10.4%-11.2% yield. If approximately $75-$90 million of annual repurchases is required merely to offset awards, cash available for true net retirement or other owners’ uses is lower.

The current price therefore embeds more than simply zero growth. It appears to discount some combination of declining normalized cash flow, capital-allocation risk, forecast skepticism, and a required double-digit yield. If ARR stabilizes and net cash remains intact, that discount can close without high growth. If owner earnings fall toward $200-$225 million, the apparent cheapness largely disappears.

Peer framing

Snowflake and MongoDB are operating competitors but poor direct valuation peers. Snowflake’s latest product revenue grew 37% with 126% net revenue retention; MongoDB’s revenue grew 30% with 122% net ARR expansion. Teradata’s Q2 revenue was flat, Total ARR grew 1%, and 2025 cloud net expansion was 108%. Higher growth-platform multiples are economically justified by expansion and ecosystem position. [S1] [S2] [S13] [S14]

The comparison also cuts against uncritical peer enthusiasm. Snowflake remained GAAP loss-making and incurred more than $400 million of quarterly stock compensation. MongoDB also incurred substantial stock compensation. Teradata’s GAAP profitability and net cash are real strengths. A decision-useful comparison is growth plus post-stock-compensation profitability, not sales multiples alone.

Oracle and Microsoft are strategically relevant but poor consolidated valuation comparables because cloud infrastructure, applications, and other operations dominate their corporate results. Their relevance is competitive distribution: they can bundle databases with infrastructure, identity, productivity software, and AI. Teradata deserves a mature specialist multiple until it proves durable ARR acceleration.

Own-history context

The current forward management-defined FCF multiple is near the low end of recent observations, but historical percentiles are unstable because the business mix, restructuring, settlement, and lease treatment changed. At the 2025 trough, expectations reflected ARR contraction and repeated misses. The current multiple should not be declared cheap solely because it resembles a prior trough.

The August 2026 decline occurred despite higher adjusted EPS and FCF guidance, indicating that the market assigned greater importance to revenue and ARR durability. That is rational for a business whose terminal value depends on retention rather than one year’s cost savings.

Scenario analysis

These are analyst estimates. Net cash is reduced modestly in the scenarios to allow for working capital, pensions, and allocation. The bear case assumes current savings cannot offset erosion indefinitely. The base case requires retention, modest expansion, and net share retirement—not peer-like growth. The bull case requires attributable product monetization and a higher-quality revenue mix.

A 30% bear, 50% base, and 20% bull weighting produces an expected value near the mid-$30s. The largest sensitivity is owner-earnings durability, followed by ARR growth and the multiple. One percentage point of operating margin is less important than whether the recurring base remains intact.

What the market gets right

The market correctly discounts Teradata for declining five-year revenue, opaque total retention, weaker cloud expansion, forecast misses, stock compensation, and unproven AI monetization. It also correctly refuses to value product announcements as revenue.

What may be mispriced

The market may underweight the removal of term debt, low customer concentration, improved GAAP margin, positive Total ARR, and the possibility that hybrid demand slows forced migration. If lease-adjusted owner earnings remain near $280 million and the share count declines, per-share value can grow without a broad competitive resurgence.

Fragile assumptions

The bullish valuation is fragile to three items: recurring margin below 65%, cloud ARR remaining flat, and repurchases failing to reduce shares. The bearish valuation is fragile if total retention proves strong, AI workloads become separately paid, or management consistently converts ARR into owner earnings above $300 million.

Verdict: Valuation provides a margin of safety only against stabilization. The headline 13% FCF yield overstates economic cash generation, but a roughly 10%-11% lease-adjusted yield remains attractive if ARR does not contract.

Variant Perception

The apparent market view is that Teradata is a legacy warehouse vendor with a stalled cloud transition, weak forecast credibility, and AI positioning that will not overcome Snowflake, Databricks, hyperscalers, and modern databases. This is an inference from the low multiple and event reactions, not a measured consensus survey. Current sell-side estimate distributions were not independently available.

Thoughtful investors have focused on why strong product commentary does not produce near-term acceleration; whether Snowflake and Databricks are taking incremental workloads; whether AI spending adds to budgets or reallocates them; how much Q4 ARR is contracted; what retention underlies guidance; and how hardware affects cash economics. [S5]

Management’s answers were informative but incomplete. It attributed Q3 revenue weakness to first-half upfront license recognition, expects most ARR growth in Q4, views hybrid demand as durable, and said new products were not yet material. Those statements define tests; they do not satisfy them.

Strongest bull case

The market may mistake a durable installed base for terminal decline. Total ARR returned to growth in 2025, margins improved, term debt was eliminated, and a 9.9% shareholder secured board refreshment. Switching costs and sovereign requirements make rapid displacement difficult. At a double-digit lease-adjusted earnings yield, modest ARR growth plus net buybacks could produce acceptable returns. OneLake and Iceberg interoperability may allow Teradata to monetize selected compute workloads without controlling all storage. [S1] [S2] [S8] [S11]

The bull case does not require Snowflake-like growth. It requires the existing base to last longer than the market expects and management to allocate cash below intrinsic value.

Strongest bear case

The apparent stabilization may be temporary. Cloud migrations could be nearing exhaustion, maintenance and subscription ARR continue to erode, and restructuring has already captured the easiest savings. Public Cloud ARR’s sequential stall at $686 million is an early warning. The 2024 guidance miss shows that management’s pipeline assumptions can fail materially. AI products may defend relevance without creating new consideration, while lease principal and stock compensation make headline FCF look better than owner economics. [S1] [S2] [S4] [S22] [S23]

Peer performance strengthens the bear case: enterprise data demand is robust, yet Teradata grows much more slowly. Its weakness cannot be explained solely by a poor industry environment. [S13] [S14]

Load-bearing assumptions

1. Total ARR must remain positive because cloud ARR can rise through migration while company economics stagnate.
2. Recurring gross margin must remain in the high-60s despite cloud and hardware mix.
3. Lease-adjusted owner earnings must remain above approximately $275 million.
4. Repurchases must reduce diluted shares after awards.
5. AI, Factory, and interoperability must improve retention or expansion within four to six quarters.

Positioning and factor context

The factor model dated September 11, 2026 shows market exposure of 1.06, positive statistical loadings to Technology, NewValue, and SmallSize, low CreditRisk exposure, residual volatility of 0.458, and R-squared of only 0.17. Sector coefficients are return exposures, not legal or economic industry classifications. Low explanatory power means broad factors explain only a small portion of the fitted variation; company-specific earnings and execution events remain important. The positive residual-momentum reading of 0.040 is small relative to residual volatility and should not drive the fundamental call. [S10]

Variant conclusion

The best variant is not that Teradata becomes an AI hypergrowth leader. It is that a sticky installed base, repaired balance sheet, and modest ARR growth support more durable owner earnings than the price implies. That view is falsified by Total ARR contraction or lease-adjusted owner earnings below $250 million.

The bear view is falsified if total retention improves, cloud ARR grows above 15%, paid AI workloads become attributable, and recurring margin holds. Until then, the market’s skepticism is substantially evidence-based.

Verdict: The investable variant is modest stabilization, not technological triumph. It is credible but requires at least two quarters of measurable confirmation.

Fact vs. Interpretation

The discipline is to keep product availability, management confidence, and engagement counts separate from paid recurring economics. GAAP settlement income, adjusted FCF, and lease-adjusted owner earnings answer different questions and should not be substituted for one another. [S2] [S5]

Verdict: Verified facts support liquidity and partial stabilization. Competitive resurgence, high ROIC, and AI monetization remain estimates or hypotheses.

Open Questions

1. What are gross and net retention for Total ARR, and how much change reflects migration, expansion, new logos, downsizing, and churn? The SEC previously requested related disclosure, and the gap remains decision-useful. [S20]

2. How much of the year-end ARR requirement is already contractually committed versus dependent on unsigned Q4 renewals or expansions? [S5]

3. What paid ARR, revenue, and gross profit are directly attributable to AI Studio, Tera, Factory, and OneLake-connected workloads? [S5] [S7] [S8] [S19]

4. What are public-cloud, subscription-software, maintenance, and hardware-rental gross margins separately? [S1]

5. How much finance-lease principal and asset acquisition represent maintenance versus growth equipment? [S1] [S2]

6. Which competitors are displaced in reported wins, and which wins are merely migrations between Teradata deployments? [S1]

7. Does OneLake integration create incremental consumption or reduce proprietary renewal leverage? [S8]

8. Will diluted shares continue to decline after the expanded equity plan and director grants? [S2] [S4] [S26]

9. Will the compensation committee avoid further target resets following operational misses? [S4]

10. What acquisition-size and leverage limits will the refreshed board observe? [S5] [S27]

11. What are Factory’s equipment returns, useful lives, supplier terms, and working-capital requirements? [S2] [S19]

12. Are consulting losses a temporary partner transition or evidence of reduced implementation influence? [S1] [S2]

13. What changed in forecasting controls after the roughly ten-point midpoint miss on 2024 Total ARR? [S12] [S22] [S23]

These are genuine missing data rather than rhetorical objections. Total-company retention is the most important because it would separate installed-base durability from deployment reclassification.

What Must Be True

Bull tests

• Total ARR durability: Total ARR must grow at least 3% year over year for two consecutive quarters. The current premise is only 1% reported Q2 growth and a 2%-4% year-end guide. Falsifier: contraction or growth entirely caused by currency. Monitor reported and constant-currency ARR each quarter. [S2] [S3]

• Cloud reacceleration: Public Cloud ARR must resume sequential growth and exceed 12% year-over-year growth by mid-2027. Falsifier: cloud ARR remains near $686 million through year-end or growth stays below 10%. Monitor ARR dollars, cloud net expansion, and migration disclosures. [S1] [S2] [S21]

• AI and Factory monetization: Management must disclose paid recurring workloads, separate pricing, or matched retention evidence. Falsifier: continued reliance on launches and engagement anecdotes four to six quarters after introduction. Monitor attributable ARR, renewal, consumption, and gross margin. [S5] [S7] [S19]

• Margin durability: GAAP recurring gross margin must stabilize near 67%-69% while non-GAAP operating margin remains at least 20%. Falsifier: recurring margin below 65% or reinvestment eliminating operating leverage. Monitor hosting cost, hardware mix, and GAAP-to-adjusted exclusions. [S1] [S2] [S3]

• Cash quality: Adjusted FCF must remain at least $330 million and lease-adjusted owner earnings above $275 million. Falsifier: adjusted FCF below $300 million or rising lease investment without corresponding rental ARR. Monitor CFO, cash capex, lease additions, principal, and rental revenue. [S1] [S2] [S3]

• Per-share compounding: Diluted shares must decline after all awards. Falsifier: annual dilution despite repurchases equal to roughly half of adjusted FCF. Monitor gross repurchases, award issuance, withholding, and diluted shares. [S1] [S2] [S4]

• Forecast rehabilitation: The company must achieve the current year-end ARR range without another material reset. Falsifier: a guidance cut or unexplained Q4 shortfall. The 2024 miss makes this a separate test rather than a routine modeling input. [S12] [S22] [S23]

Bear tests

• Competitive erosion: The bear case requires maintenance and subscription losses to exceed cloud additions, causing Total ARR contraction. Bear falsifier: Total ARR accelerates above 5% with improved disclosed retention. Monitor deployment-level ARR and recurring gross profit. [S1] [S2]

• Peak self-help: The bear case requires margins to plateau as restructuring savings are exhausted. Bear falsifier: revenue resumes growth while GAAP and adjusted margins remain stable or rise. Monitor SG&A, R&D, headcount, and incremental margin. [S1] [S22]

• AI irrelevance: The bear case requires new products to remain bundled or non-incremental. Bear falsifier: separately identified AI or Factory ARR renews and expands at attractive gross margin. Monitor contracts rather than engagement counts. [S5] [S6] [S7] [S19]

• Cash-flow illusion: The bear case requires leases, dilution, and working capital to consume headline FCF. Bear falsifier: lease-adjusted owner earnings exceed $300 million and diluted shares fall without added leverage. Monitor the complete cash and share bridge. [S1] [S2]

• Capital-allocation failure: The bear case strengthens if net cash funds an expensive acquisition or price-insensitive repurchases. Bear falsifier: the company remains net cash, buys below conservative value, avoids material goodwill expansion, and does not reset incentives. [S2] [S4] [S5]

The decisive monitoring sequence is Q3 Total ARR, the remaining Q4 requirement, year-end cloud ARR, recurring gross margin, adjusted FCF, finance-lease principal, and diluted shares. Product launches should change the thesis only when those measures change. [S2] [S3] [S5]

Verdict: The bullish case requires measurable stabilization rather than heroic share gains. The bearish case requires proof that current cash flow is a temporary product of cost cuts, accounting timing, or underinvestment.

Primary linked evidence: 2025 Form 10-K, Q2 2026 Form 10-Q, Q2 2026 earnings release, 2026 proxy statement, Snowflake July 2026 Form 10-Q, and MongoDB July 2026 Form 10-Q.

Public source appendix

• S1: Teradata 2025 Form 10-K — primary filing; published 2026-02-27; Items 1, 1A, 7 and 8; financial statements; ARR, revenue, leases, obligations, stock compensation, segments, repurchases and geographic notes
• S2: Teradata Form 10-Q for the quarter ended June 30, 2026 — primary filing; published 2026-08-05; Financial statements and MD&A; ARR, SAP settlement, debt repayment, leases, repurchases, foreign cash and litigation
• S3: Teradata Q2 2026 earnings release — primary company release filed with SEC; published 2026-08-04; Q2 results, Q3 and full-year outlook, cash-flow guidance and non-GAAP reconciliations
• S4: Teradata 2026 proxy statement — primary proxy filing; published 2026-03-25; Ownership, compensation metrics, incentive results, goal resets, burn rate, overhang and board refreshment
• S5: Company Financials — Teradata Q2 2026 earnings-call transcript — management transcript via Company Financials; published 2026-08-04; August 4, 2026 prepared remarks and Q&A on revenue timing, Q4 ARR weighting, retention, new products, hardware, consulting and M&A
• S6: Company Financials — Teradata Q1 2026 earnings-call transcript — management transcript via Company Financials; published 2026-05-05; May 5, 2026 prepared remarks and Q&A on hybrid customers, upfront licenses, AI services and product materiality
• S7: Teradata introduces the Autonomous Knowledge Platform — primary product announcement; published 2026-05-07; Product capabilities, elastic compute, AI Studio, Tera and deployment description
• S8: Teradata brings enterprise AI to Microsoft OneLake — primary product announcement; published 2026-09-02; OneLake integration and Iceberg interoperability description
• S9: Company Financials — TDC historical market prices — market data via Company Financials; publication date unavailable; Split-adjusted daily OHLC history through September 11, 2026; closing and intraday extrema distinguished
• S10: The factor model — TDC snapshot — quantitative diagnostic; published 2026-09-11; Dated exposures, residual signals and diagnostics
• S11: Teradata cooperation agreement with Lynrock Lake — primary filing; published 2026-02-11; Item 1.01, board-refreshment commitments, standstill and voting provisions
• S12: Teradata 2024 Form 10-K — primary filing; published 2025-02-21; Business, MD&A, 2024 ARR outcome, financial statements, leases, repurchases and risk factors
• S13: Snowflake Form 10-Q for the quarter ended July 31, 2026 — primary peer filing; published 2026-09-04; Product revenue, net revenue retention, large customers, gross margin, GAAP losses and stock compensation
• S14: MongoDB Form 10-Q for the quarter ended July 31, 2026 — primary peer filing; published 2026-09-01; Revenue, Atlas mix, net ARR expansion, margins and stock compensation
• S15: Teradata director Michael Gianoni Form 4 — primary insider filing; published 2026-02-25; Table I, February 24, 2026 sale of 25,000 shares at $29.965
• S16: Teradata 2022 Form 10-K — primary filing; published 2023-02-24; 2022 and comparative financial statements, R&D and cash flow
• S17: Teradata 2023 annual report — primary filing; published 2024-03-27; 2023 financial statements, MD&A, ARR, leases and comparative data
• S18: Teradata 2021 Form 10-K — primary filing; published 2022-02-25; Business, 2021 financial statements, headcount, revenue and capital allocation
• S19: Teradata announces Factory for on-premises and sovereign AI — primary product announcement; published 2026-05-19; Factory architecture, Dell and NVIDIA infrastructure, sovereignty and hybrid deployment
• S20: SEC comment letter concerning Teradata retention disclosures — regulator correspondence; published 2022-09-21; Comment 1 concerning cloud net expansion and monitoring of other recurring revenue
• S21: Teradata Q1 2026 earnings release — primary company release filed with SEC; published 2026-05-05; Q1 Total ARR and Public Cloud ARR, revenue, margins and outlook
• S22: Teradata Q2 2024 earnings release — primary company release filed with SEC; published 2024-08-05; Q2 ARR, reduced 2024 guidance, restructuring scope, expected savings and severance
• S23: Teradata 2023 results and initial 2024 outlook — primary company release filed with SEC; published 2024-02-12; Initial 2024 Total ARR and Public Cloud ARR guidance
• S24: Teradata Q3 2025 earnings release — primary company release filed with SEC; published 2025-11-04; Q3 ARR, revenue, adjusted free cash flow and maintained outlook
• S25: Teradata 2025 results release — primary company release filed with SEC; published 2026-02-10; Year-end Total ARR, Public Cloud ARR, revenue, free cash flow and 2026 outlook
• S26: Teradata August 2026 director Form 4 filings — primary insider filing; published 2026-08-18; August 14, 2026 restricted-unit award under the Director Compensation Program; representative of contemporaneous director filings
• S27: Teradata appointment of Bernd Leukert to the board — primary filing; published 2026-08-03; Item 5.02; board expansion and director appointment effective August 1, 2026