Predictive AI Statistics 2026: Adoption, Spending and Accuracy

Gartner forecasts spending on AI platforms for data science and machine learning at $26,444 million in 2026, up 36.3% on the year. That platform category is where most fraud scores, demand forecasts, and risk models get built and retrained. Predictive AI statistics in the same forecast show foundation GenAI models growing faster, from 11,438 to 23,356 million, but they still sit below the predictive toolset in the same forecast.

Adoption is harder to pin down than spending. A Federal Reserve note found adoption stood at about 18% of firms at the end of 2025 in Census data. The Atlanta Fed’s executive survey, by contrast, put an employment-weighted firm AI adoption rate of around 78% in the same economy. The sections below cover spending, adoption, fraud detection, forecasting accuracy, healthcare and maintenance, with every figure traced to a primary document.

Key Takeaways

• Gartner expects spending on GenAI models to grow 117%, while AI platform spending will rise 36.9% in 2026.
• Census researchers found 18% of firms used AI in a business function, rising to 32% on an employment-weighted basis.
• Mastercard’s initial modelling shows AI enhancements boost fraud detection rates on average by 20% and as high as 300% in some instances.
• The GraphCast preprint reports that GraphCast significantly outperforms the most accurate operational deterministic systems on 90% of 1380 verification targets.
• In two county emergency departments, the ESPMv1 had a sensitivity of 14.7% for sepsis within a 6-hour window.
• A properly functioning predictive maintenance program can provide a savings of 8% to 12% over a program that utilizes preventive maintenance alone.
• Gartner predicts over 40% of agentic AI projects will be canceled by the end of 2027.

Editor’s Choice

• Global corporate AI investment: $581.7 billion in 2025, up 130% from the prior year.
• AI platforms and models spending: projected to total $64 billion in 2026, up 63.4% from $39 billion in 2025.
• AI-optimized IaaS spending: projected to grow 96% through 2026, reaching $42 billion.
• Visa transaction volume: Visa processes over 269 billion transactions each year.
• Visa fraud prevented: Decision Manager screened 3.2 billion transactions and prevented an estimated $33 billion in potential fraud losses in 2023.
• Walmart Self-Healing Inventory: saved Walmart more than $55 million.

Predictive AI Statistics on Market Size and Spending

• Worldwide end-user spending on AI models and platforms is projected to total $64 billion in 2026, per Gartner’s July 2026 forecast.
• Data science and machine learning platforms account for about 41.2% of that 2026 total.
• The predictive-platform segment outspends foundation GenAI models by $3,088 million in 2026.
• Gartner analyst Arunasree Cheparthi said Enterprise AI budgets are coming under greater scrutiny, with increased focus on usage efficiency, cost control and measurable outcomes.
• Domain-specific language models (DSLMs) and specialized models are forecast to grow 210% in 2026, the fastest segment in the table.

Source: Gartner, July 2026

About This Data

• Compiled from 25 sourced figures across 22 documents: 21 from primary publications (Gartner, Census Bureau, Federal Reserve, FDA, ECMWF, PNNL, Mastercard, Visa and Walmart publications) and 4 from research (Stanford HAI, an arXiv preprint and a peer-reviewed clinical study).
• Source publication dates run from February 2024 to September 2026; one maintenance benchmark traces back to a 2010 federal guide and is labeled as such.
• Only primary documents, official statistics, and original research qualified; market-research resellers and news summaries were excluded.
• Figures are reviewed on a rolling basis and updated when the underlying sources publish new editions.

Don’t chase tech news. We track it for you.

One weekly briefing with the launches, AI developments, and breaches that matter. No filler.

Email address

Subscribe Now

Read by pros at Fortinet, TSMC, Barclays, and Deloitte.

What Is the Difference Between Predictive AI and Generative AI?

Predictive AI scores or forecasts an outcome from historical data, such as a fraud risk or tomorrow’s temperature. Generative AI produces new text, images, or code. Gartner tracks the two budgets separately: AI Platforms for Data Science and Machine Learning sit in a different line from Foundation Generative AI Models in its 2026 forecast.

US Business AI Adoption Rates

• Census Bureau survey data show overall AI usage hovered between 17% and 20% from December 2025 to May 2026.
• Looking ahead, between 20% and 23% of businesses are expected to be using it in the next six months.
• Small firms lag: less than 20% of firms with four or fewer employees reported using AI.
• In retail, around 14% of businesses currently use AI, and about 17% expect to in the next six months.
• About 42% of businesses in the Information sector and roughly 39% in Finance and Insurance expected to use AI in their business functions over the next six months.
• A Census working paper found use rates reaching 50% to 60% (60% to 70%, employment-weighted) for very large firms in the Information, Professional Services, and Finance sectors.

Recent Developments

• September 2026: Gartner predicts that by 2029, 30% of employees laid off due to replacement by AI will need to be rehired, often at a significantly higher cost.
• August 2026: The market for securing AI is projected to reach almost $4.8 billion in 2027, a 68.7% increase over 2026, according to Gartner.
• August 2026: Gartner found that routing a task to an agentic reasoning model increases provider inference costs by at least five times compared with a basic chatbot interaction.
• August 2026: Gartner expects global spending on inference ($23.3 billion) will surpass that of training ($19 billion) in 2026.
• July 2026: Gartner’s platform forecast put the Total Market at $39,311 million in 2025 and $64,252 million in 2026.
• May 2026: ECMWF put AIFS v2 into operation, introducing ECMWF’s first data-driven wave and snow cover forecasts.

Why AI Adoption Estimates Range From 18% to 78%

• A Federal Reserve note cites research in which Crane, Green, and Soto (2025) examined 16 such surveys and found that, as of mid-2024, point estimates of work-related AI adoption rates ranged from about 5 to 40%.
• The Real-Time Population Survey shows work-related GenAI adoption reported in the RPS stands at about 41% of the workforce.
• The same survey puts non-work-related usage at about 50% of the population as of November 2025.
• Prior to the question revision, the adoption rate had grown by 68% (3.9 percentage points) over the prior year.
• In the Atlanta Fed survey, employment weighting approximates the share of the labor force working at firms that have adopted AI.
• Census researchers report that in 23% (41%, employment-weighted) of firms, workers use AI in work-related tasks.

Source: Federal Reserve FEDS Note, April 2026

Why it matters: A headline adoption rate means little without its unit. The Federal Reserve note puts Census firm-level adoption at about 18% of firms, while the Atlanta Fed survey’s SBU estimates an employment-weighted firm AI adoption rate of around 78%. Large employers adopt first, so weighting by workers inflates the rate.

Which Business Functions Use AI Most

• Among AI adopters, 57% of users integrate AI in three or fewer business functions.
• Writing, document analysis, and information search are the leading Generative AI uses in tasks, though 65% of firms limit use to three or fewer tasks.
• The Census team expects adoption is expected to reach 22% within six months of the Nov 2025 to Jan 2026 reference period.
• The leading functions lean toward writing and search, a pattern our AI agent adoption data also shows.

Global AI Investment Statistics

• Stanford’s AI Index reports private investments reached $344.7 billion, an increase of 127.5% from 2024.
• Its investments ($285.9 billion) were 23.1 times greater than those of the next-highest country, China, for the United States.
• Generative AI reached 53% population adoption within three years.
• Adoption outliers include Singapore (61%) and the United Arab Emirates (54%), while the U.S. ranks 24th at 28.3%.
• The estimated value of generative AI tools to U.S. consumers reached $172 billion annually by early 2026.
• Round-by-round detail on the largest checks sits in our AI funding rounds tracker.

AI Infrastructure and Inference Spending

• Gartner forecasts Inference-Driven AI Infrastructure Spending to Grow 55% in 2026.
• The inference share of AI-optimized IaaS spending is set to reach 59% in 2027.
• Total IaaS spending moves from 222,170 million in 2025 to $287,347 million in 2026.
• AI-optimized services make up about 14.7% of total IaaS spending in 2026.
• AI inference costs per agentic workflow will increase more than fivefold through 2028, per Gartner.

Predictive AI in Fraud Detection

• Mastercard’s Decision Intelligence (DI) – a real-time decisioning solution – already helps banks score and safely approve 143 billion transactions a year.
• Its newer model will scan an unprecedented one trillion data points to predict whether a transaction is likely to be genuine or not.
• Scoring happens fast: in less than 50 milliseconds, this technology improves the overall DI score.
• Visa’s Decision Manager rates each transaction from 0 (low risk) to 99 (high risk).
• For active users, Decision Manager has helped reduce manual reviews by 25% or more.

Source: Mastercard press releases (February and May 2024), Visa (2023 data)

What Are Examples of Predictive AI?

Common examples include card-fraud scoring, sepsis early-warning alerts, weather forecasting, and inventory rebalancing. Mastercard’s model doubles the detection rate of compromised cards, ECMWF runs data-driven forecasts in operation, and Walmart uses Self-Healing Inventory to reroute overstock before it becomes waste.

Payment Fraud Threats and AI Returns

• Mastercard’s 2025 research with Financial Times Longitude found that 42% of issuers and 26% of acquirers have saved more than $5 million in fraud attempts over the past two years thanks to AI.
• 83% of industry leaders say AI has reduced false positives and churn.
• On the loss side, organizations lost an average of $60 million to payment fraud in the past year.
• Synthetic identities and deepfakes also run through our cybersecurity threat data, which tracks the attack side of the same problem.

Predictive AI Accuracy in Weather Forecasting

• The GraphCast model, described in an arXiv preprint, predicts hundreds of weather variables, over 10 days at 0.25-degree resolution globally, in under one minute.
• ECMWF reports that data-driven models outperform leading physics-based global numerical weather prediction (NWP) models in many of the standard forecast scores.
• ECMWF’s first AIFS version was trained on a subset of the ERA5 reanalysis for 1979 to 2018.
• AIFS v2 made eleven wave-related variables available, and the AIFS Single v2 wave component shows substantial improvement in the medium-range skill of wave forecasts compared to the physical model, IFS Cycle 50r1.
• Stanford’s AI Index notes that for the first time, AI ran a full weather forecasting pipeline end-to-end.

Source: GraphCast arXiv preprint (2023), ECMWF Newsletter 178

Worth noting: ECMWF’s own description of AI weather models carries a dependency that headline accuracy scores hide. The data-driven systems rely on traditional NWP analyses as initial conditions when producing a forecast, and the AIFS at a resolution of one degree is still behind the operational IFS configuration for surface fields such as 2 m temperature.

Predictive AI Accuracy in Healthcare

• A 2023 study at two county emergency departments evaluated the predictive characteristics of the ESPMv1 for 145 885 encounters.
• The ESPMv1 BPA alert was present in 7183 (4.9%) encounters, of which 2253 had sepsis.
• The researchers found that Providers were alerted with a median lead time of 0 minutes.
• Stanford’s AI Index found that in a review of more than 500 clinical AI studies, only 5% used real clinical data.
• The FDA says its device list is not a comprehensive resource of AI-enabled medical devices, and the FDA will explore methods to identify and tag medical devices that incorporate foundation models.
• Rules for clinical AI differ by market, and our AI regulation tracker follows them jurisdiction by jurisdiction.

Key finding: The county emergency department study found the Epic model only alerted providers in half of the cases prior to sepsis occurrence. The authors concluded that the ESPMv1 alert is adding little assistance to physicians identifying sepsis, a reminder that vendor-reported accuracy needs independent validation at each site.

How Accurate Is Predictive AI?

Predictive AI accuracy depends on the domain, the data, and the metric chosen. In weather, GraphCast significantly outperforms the most accurate operational deterministic systems on 90% of 1380 verification targets. In sepsis alerts, the Epic model had a sensitivity of 14.7% and a specificity of 95.3% in two county emergency departments.

Predictive Maintenance and Supply Chain Savings

• According to PNNL’s best-practice summary, a well-orchestrated predictive maintenance program will all but eliminate catastrophic equipment failures.
• Facilities starting from reactive maintenance could easily recognize savings opportunities exceeding 30% to 40% (FEMP 2010).
• Those benchmarks come from a 2010 federal guide, so they describe condition monitoring and vibration analysis rather than modern machine learning models.
• Walmart says produce is sorted by predictive AI in Costa Rica before sunrise.
• These intelligent systems are already live across markets like Costa Rica, Mexico, and Canada, predicting demand and rerouting inventory.

Source: PNNL O&M Best Practices, citing FEMP 2010

Agentic AI Adoption Forecasts

• Gartner expects 40% of enterprise applications will be integrated with task-specific AI agents by the end of 2026, up from less than 5% today.
• In its best case, agentic AI could drive approximately 30% of enterprise application software revenue by 2035, surpassing $450 billion, up from 2% in 2025.
• Gartner estimates only about 130 of the thousands of agentic AI vendors are real.
• Gartner also predicts at least 15% of day-to-day work decisions will be made autonomously through agentic AI by 2028.
• 33% of enterprise software applications will include agentic AI by 2028, up from less than 1% in 2024.

AI Agent and Model Capability Benchmarks

• Stanford’s AI Index shows the success rate of agents handling real-world tasks improved from 20% in 2025, according to Terminal-Bench.
• Cybersecurity agents improved from 15% in 2024.
• On other tasks, AI lags behind, including learning from video, generating video that is coherent and realistic, telling time, managing multiple-step planning, and conducting financial analysis.
• Benchmark leaders change quickly, and our recent model releases tracker logs each capability jump as it lands.

Is ChatGPT Predictive or Generative AI?

ChatGPT is generative AI: it writes new text in response to prompts, even though the underlying model works by predicting the next word. Business use follows that profile, with Census researchers finding that Writing, document analysis, and information search are the leading Generative AI uses in tasks.

Conclusion

Gartner’s forecast of $26,444 million for data science and machine learning platforms this year shows predictive tooling still carries more spend than foundation GenAI models. Adoption numbers need more care: the Federal Reserve places firm-level use at about 18% of firms, while employment-weighted survey data reaches around 78%. Buyers, analysts, and regulators get the most from predictive AI when they ask which unit, which metric, and which site produced a figure.

The next test is independent validation, which the sepsis study shows can diverge sharply from vendor results. Workforce effects remain open, as our AI employment statistics show.