The demand is certain. The data is not ready.
The headline number from an MIT Technology Review piece is unanimity: within two years, 100% of respondents plan to be using agentic AI, and 69% expect to use it widely.1 The same piece reports the catch. In organizations it labels 'data laggards', AI access to company data falls to 30% or less.1 Agents that cannot see most of a company's information cannot do much of its work.
For an investor, the money flows to whoever removes that bottleneck: chips and software to run agents, and tools to govern the data they touch. The trouble is that the article is a weak foundation. Via News measured the reliability of that source, and 0% of the 11 claims we checked held up.1 That is a small sample, but it is a poor record. Treat the survey percentages as a signal of mood, not as audited fact. The Nvidia numbers below are different: they come from SEC filings and are checked.
What the verified numbers say about Nvidia
Nvidia is the clearest listed company in this story, though our dossier carries only a few of its financial lines. Revenue is not among them, so we cannot show profit margins. We can show how much it costs the company to deliver what it sells. Its cost of revenue was $16.6 billion in fiscal 2024, $32.6 billion in fiscal 2025 and $62.5 billion in fiscal 2026.2 That is roughly 3.8 times the fiscal 2024 level in two years. For scale, $62.5 billion is nearly four times what the same line cost two years earlier. We have no outside yardstick in our data, so we will not invent one.
The latest quarter points the same way. Cost of revenue in Q1 2027 was $20.5 billion, against $17.4 billion in Q1 2026, $5.6 billion in Q1 2025 and $2.5 billion in Q1 2024.2 That is about 18% growth over the past year, and a much larger climb over three years. A rising cost of revenue on its own can mean growth or a squeeze. Without revenue in the dossier, it only tells us the business is much bigger than it was.
Cash tells a less steady story. Year-end cash was $7.3 billion in fiscal 2024, $8.6 billion in fiscal 2025 and $10.6 billion in fiscal 2026.3 Quarter by quarter it moves around: $15.2 billion in Q1 2026, $11.6 billion in Q2 2026, $11.5 billion in Q3 2026 and $13.2 billion in Q1 2027.3 We cannot say from this data why it moves, and we will not guess. Earnings per share is also in the dossier: $11.93 in fiscal 2024, $2.94 in fiscal 2025 and $4.90 in fiscal 2026.4 Those figures are not on a comparable basis as presented, and we do not read a trend into them.
Do you own it? The dossier does not give Nvidia's weighting in any index fund, so we will not say. Check your own fund's holdings list.
The entity graph shows why Nvidia sits in the agent story. It develops NeMo Guardrails and the NeMo Agent Toolkit, and counts Mount Sinai Health System, the State of Alaska Legislative Affairs Agency and Yum! Brands as customers; Advanced Micro Devices, Inc. competes with it.5 The guardrails product matters most to this theme, because trust in agents is the constraint.
Startups are selling against payroll, not software budgets
The startups in this wave size their markets by labor spend. Penguin AI's Glenn Herzberg says the company defines its market as administrative labor spend rather than the healthcare IT software budget. He added: "US Healthcare Administration runs about a trillion dollars a year, about a quarter of the total health spend, and the published estimates put around $570 billion of that in work that has no effect on health outcomes."6 That is the company's own pitch. We have not verified the $570 billion estimate.
Covecta's Ben Thomas makes the same move in banking: "Covecta’s total addressable market (TAM) is tens of thousands of financial institutions globally, and for them we are not just disrupting their software budget but their labor budget as well."7 He says the company serves banks, specialist lenders, building societies, credit unions and private credit firms, currently in the US and UK.7 Maisa AI's David Villalon describes a market of process automation at regulated industries, for tasks that must be auditable, reproducible and hallucination resistant.8
Investors should note the shape of these claims. A labor-budget pitch makes the market look huge, but it only pays if customers trust the agent enough to replace people. Maisa's own list of requirements, auditable, reproducible and hallucination resistant, is a list of trust problems. All three are executives talking about their own companies, and none of these interviews gives revenue or funding figures.
Incumbents are buying governance first
Large companies are spending on control as much as capability. Manulife and Microsoft announced a five-year agreement. Manulife will adopt Microsoft's Frontier Suite, deploy Microsoft Agent 365 and expand Microsoft 365 Copilot to more than 30,000 employees.9 Shamus Weiland called the partnership "a critical enabler of Manulife's continued evolution into a truly AI-driven organization."9
Box announced controls for AI agents on enterprise content: agent guardrails, oversight of third-party agent activity, prompt injection detection and agent classification-based access policies.10 Nomura Research Institute's Tatsutoshi Murata said the company expects Box to "provide the administrative features needed to safely leverage this new era of AI."10 Both pieces come from company press releases. Via News measured this news source at 57% of 4,954 checked claims holding up.910 Read them as what the companies want investors to believe, not as results.
What to watch
- Whether Nvidia's cost of revenue keeps climbing, and whether revenue and cash keep pace. Cash has swung between quarters.23
- Whether 'data laggards' close the gap beyond 30% access, the share the MIT piece cites.1
- Whether the startups publish revenue, funding or customer results, rather than market-size claims.67
- Whether governance products, such as Box's agent controls, show up in customer numbers.10
Our reading, offered as Via News analysis and not fact: demand for agents looks settled, but the evidence that data and trust are the limiting factors is thinner than the headlines suggest. The strongest checked numbers here are Nvidia's filings, and they show spending, not whether agents pay off.
