Is the AI market mostly PR, or is there deeper math at work behind the headlines? A market analysis from AI software company Linkmate, formerly Linkomo, argues that the answer lies in a mix of loan dynamics, capital spending, and measured business demand.

The loan math behind AI hype

According to The State of the Agentic Market, AI adoption is entering a later stage of maturity. In practical terms, Linkmate says the market is moving away from treating AI as a magical silver bullet and toward using it as a tool for solving ordinary business problems.

The report argues that AI is often described as a major disruptor in the labor market, but that it is also being used as a scapegoat. In Linkmate’s view, a major factor behind labor market weakness is the combination of COVID-era loans and Federal Reserve rate increases.

In 2019, large technology companies began reassessing their office footprints, asking why they should pay for four office floors in New York or Silicon Valley if they no longer needed them. That shift became an early cost-cutting measure. The lockdown period also increased corporate borrowing, as companies took on more debt at lower rates.

Between 2019 and 2020, the Federal Reserve rate reached 0.05%, making borrowing cheaper than it had been in more than 70 years. Linkmate cites Federal Reserve statements and FRED data showing that corporate loans in 2020 were 0.25%, which the analysis describes as meaning every borrowed dollar was worth only 0.25 cents, not 25 cents.

The analysis says corporate borrowing was not used conservatively. In normal periods, corporations tend to take out smaller loans and repay them quickly, but that did not happen here. Over a few years, total corporate loans rose from about $2 trillion to $3.1 trillion in the first half of 2020. The figure then fell by $0.6 trillion, or $600 billion, to $2.4 trillion in 2022, before rising again.

Linkmate says each market cycle has brought new ways for companies to increase profits and reduce pressure from debt. In 2023, OpenAI introduced a large language model and interactive chatbot that could answer questions dynamically. According to Linkmate, this created the expectation that companies could replace part of their cost base, cut significant portions of their workforce, and improve profits.

That expectation, the analysis says, helped fuel investor optimism and a wave of layoffs across major technology companies, including Oracle, Google, Amazon, Payoneer, Intel, HP, Meta, Fiverr, xAI, Salesforce, CISCO, and Peloton. TechCrunch published a full list for 2025. Linkmate says layoffs have risen every year since 2023, supporting the view that the AI market is at the peak of Gartner’s hype cycle.

Where AI sits in Gartner’s hype cycle

Gartner’s market hype model argues that technologies pass through stages of maturity. After the peak of inflated expectations, the next phase is usually a reassessment of the market and a more sober look at how the technology actually applies.

Linkmate says one of AI’s most promising uses is solving real-world equivalents of NP-complete tasks. The report points to applications such as protein-chain analysis, discovering treatments for old diseases, advancing cancer research, saving lives through unconventional methods, identifying new materials, and making medicine cheaper. While AI is not a silver bullet, the analysis says machine learning and large language models can handle tasks that are too cumbersome for conventional methods.

Once the market moves beyond inflated expectations and into a more sober phase, Linkmate argues, those practical use cases begin to emerge and are recognized as direct solutions to existing problems.

The analysis also says the market is confronting the idea that people are still cheaper than AI. A trend described as “tokenmaxxing,” or maximizing token use, is presented as an example of the peak of inflated expectations — the belief that AI can replace human labor in most areas. In practice, Linkmate says, the bill for AI tokens is often higher than the average salary of a software engineer.

Microsoft is quoted as saying: “Using tech is more expensive than paying human employees.” The report also cites Uber, where executives publicly decided to cut the HR department after realizing they had burned through their AI budget in four months. Nvidia’s active vice president is also cited as saying AI costs “far more than humans.”

Linkmate points to Goldman Sachs projections suggesting AI use could rise to 120 quadrillion tokens by 2030. Bank of America, in a separate bullish forecast, said AI spending could reach $155 billion by 2030.

What the numbers say about AI use

Linkmate says comparing SEC filings with public claims provides a clearer picture. The report identifies three indicators that, in its view, show AI hype is real and measurable:

• Average PJM power prices, or price per MW-day, have risen 833%.
• HBM memory has been sold out ahead of demand.
• Projected hyperscaler capex for AI in 2026 is $660 billion.

The analysis says there is a gap between promised capex and actual free cash flow reported in SEC filings. Amazon and Google are described as being in negative cash flow, while Microsoft is said to have positive numbers.

Linkmate also highlights what it calls a worrying pattern around OpenAI. According to the report, the projected annual recurring revenue milestone for 2026 is $40 billion, while real annual operational loss is negative $20 billion in 2025 and projected at negative $14 billion in 2026. The analysis says this gap between fundamentals and disclosed data could become a problem if real-world AI use cases do not materialize.

The current market narrative, Linkmate says, often frames AI mainly as an assistant for reading emails and making summaries from Google search. But the report argues that the numbers indicate demand is real, while the strongest applications are likely to be found in solving complex real-world tasks. That distinction matters because the market’s next phase is likely to depend less on headline claims and more on whether those use cases can move from pilot projects into repeatable business operations.

The analysis says adoption metrics are in place, along with venture capital spending of at least $510 billion. It adds that nearly 80% of global VC money in Q1 2026 went to AI startups, equal to roughly $242 billion out of $300 billion.

Linkmate says market estimates place the 2026 AI market at between $375 billion and $0.9 trillion, with a projected CAGR of 14% to 39%.

The report concludes that demand appears real across the board, but that supply of genuine product-market-fit offerings remains limited. In Linkmate’s view, the AI market could enter a correction phase in which weaker players are replaced by companies offering practical use cases for real-world markets. The firm’s best estimate places that possible correction between 2027 and 2028 if no widely applicable real-world use case emerges.