Markets
Nearly every second venture dollar in the U.S. went to a single category in 2024: AI and machine learning took 46.4% of deal value, according to the PitchBook-NVCA Venture Monitor PitchBook-NVCA Venture Monitor. The same category accounted for only 28.7% of completed deals, the Monitor reported, meaning the money was concentrated in unusually large checks rather than spread evenly among founders PitchBook-NVCA Venture Monitor. The venture market still calls AI a software category, which is convenient in much the same way calling an aircraft carrier a boat is convenient.
The concentration shows up fastest at the top. OpenAI raised $40 billion at a $300 billion post-money valuation, a single financing carrying a price tag comparable with a large public software company OpenAI. Anthropic raised $3.5 billion at a $61.5 billion post-money valuation, putting its price near Salesforce's annual revenue before Anthropic had anything resembling Salesforce's distribution machine Anthropic. These are not oversized software rounds. They are bets on model control, compute access and the possibility that a small group of platforms will collect tolls from the rest of the application economy.
Traditional cloud companies are being priced against retention, sales efficiency and free cash flow, the old disciplines, freshly rediscovered. AI companies are being priced against prospective market ownership. Salesforce produced $37.9 billion of fiscal-year revenue with 9% annual growth, adding several billion dollars of business while expanding at a rate venture investors would regard as pedestrian Salesforce FY2025 Results. Proven software gets a spreadsheet. AI gets a story about civilization.
The capital is not entirely detached from demand. Enterprise spending on generative AI reached $13.8 billion in 2024, up from $2.3 billion the prior year, corporate buyers expanding the category by roughly the annual revenue of a sizeable public software vendor in twelve months, according to Menlo Ventures Menlo Ventures. U.S. private investment in AI reached $109.1 billion during 2024, close to twelve times China's total, the Stanford AI Index reported Stanford AI Index. Real budgets have arrived, and the best AI products compress labor or unlock workflows that conventional SaaS vendors spent years promising to transform.
Some of that software earns its price. The trouble starts when investors use the demand to suspend ordinary distinctions between an application, a feature and an infrastructure company. A coding assistant with fast adoption may deserve a premium, but its margin structure depends on inference costs, model-provider bargaining power and whether Microsoft bundles the same capability into an existing seat. A vertical SaaS vendor owns workflow and customer history. An AI wrapper may own little beyond prompts, latency tuning and this quarter's distribution trick.
Private SaaS valuation benchmarks, meanwhile, remain terrestrial. SaaS Capital estimated valuation multiples of 4.8 times recurring revenue for bootstrapped private SaaS companies and 5.3 times for equity-backed peers, putting a business with $20 million of annual recurring revenue near a $100 million valuation rather than the billion-dollar club SaaS Capital. That company may have excellent retention, positive cash flow and customers who would complain loudly if it vanished.
It still loses the partner meeting to the agent demo.
Venture firms are structurally encouraged to widen the gap. Large funds need large outcomes, large rounds produce ownership quickly, and consensus makes career risk easier to share. Deal-sourcing systems now rank themes, relationships and founder signals with industrial efficiency, but a better technology stack does not cure a herd instinct. It lets the herd update Salesforce faster.
Conventional cloud founders face three paths under that pressure: add AI language to the pitch, accept lower prices, or build with less capital and wait for buyers to remember that predictable recurring revenue has uses. Some will benefit. Lower entry valuations improve forward returns, and disciplined operators face fewer subsidized competitors chasing identical accounts.
Marking every conventional SaaS company down while marking every AI company up is not sophistication. This is a mistake. Model costs fall, features diffuse, and enterprise procurement eventually asks the rude question venture markets postpone: who keeps paying?
The spreadsheet always gets its turn.