Andreessen Horowitz's Growth team has released its State of Markets, September 2026 1, an ambitious, 90-page examination of how technology has restructured global capital markets, reshaped the investment landscape, and ignited one of the largest infrastructure build-outs in history. The report is characteristically direct, data-dense, and unsparing in its conclusions.
Tech Is No Longer a Sector. It Is the Economy.
The report opens with a thesis that sets the tone for everything that follows. "Tech has gone from a 'cottage industry' to the prime mover in the global economy," writes a16z. "At this point, every company is, in some sense, a tech company -- you'd be hard-pressed to find a truly pen n' paper business." The numbers bear this out: tech now represents roughly 35% of the Global 100 by count and approximately 60% by market cap. It accounts for nearly half of S&P 500 profits, has delivered earnings growth roughly 3.8 times that of non-tech since 2010, and contributed approximately 76% of S&P 500 earnings growth over that same period. Tech now accounts for 55% of all U.S. capital spending. The conclusion is unambiguous: "Tech has left the cottage. Tech is the Everything Cycle now, and it's not looking back."
Yet despite new all-time highs in the market, a16z is careful to note this is not a bubble. "The market keeps setting new ATHs, but it's profits, not prices doing the work." With 93.6% of S&P 500 companies meeting or beating EPS estimates, a historic share, and tech trading at P/E multiples approximately 20% below their five-year averages, the valuation case is grounded in earnings, not speculation. "Hard to call it a 'Bubble,'" the report states plainly.
The Capex Super-Cycle: Big, and Getting Bigger
The report's most consequential section concerns AI infrastructure spending. "We'll Let You In On A Little Secret: AI Capex is Big. Really Big." Hyperscaler capital expenditures, which stood at $416 billion in 2025, are expected to reach $777 billion in 2026 and surpass $1 trillion annually from 2027. "Every year the street tries to forecast capex, and every year that forecast has fallen short."
The rationale for this investment is straightforward: demand for AI services is overwhelming supply. "Hyperscalers keep saying the same thing: Demand > Supply." Crucially, a16z argues this demand is durable. "AI infra is arguably the only pro-cyclical impulse firing in the economy right now." And the cost of building is being justified by return on invested capital well above borrowing costs, with free cash flow compression expected to resolve around 2028.
The demand mechanics are themselves noteworthy. "It's Jevons Paradox in real-time" -- as AI gets cheaper, consumption scales faster than costs fall. Token demand on OpenRouter alone has grown 27 times year-over-year, doubling monthly, and now stands at 426.2 trillion weekly tokens in September 2026. Agentic workflows have now surpassed human token usage. "Cheaper tokens have ushered in the Age of Agents," with agentic tokens reaching 7.3 trillion, a 14-fold increase, surpassing human token consumption as of February 6, 2026. OpenAI and Anthropic together are approaching approximately $150 billion in annualized revenue by Q3 2026, and "Labs have added more revenue in 2026 than all of public software combined."
Atoms Are So Back
One of the report's more counterintuitive arguments is the "bits to atoms" rotation. AI is not merely a software story. "Atoms Are So Back -- AI has led to a broader rotation from bits to atoms. We've entered a new cycle of real assets, and while tech is running through all of it, capital-light bits is no longer the (only) name of the game. Another Age of Atoms has arrived." The report cites the $90 trillion in global infrastructure investment needs projected through 2040, and notes that Waymo is now completing 500,000 weekly robotaxi rides. The winning trade in this environment has run through the physical supply chain: "The net result is the AI Supply Chain has been the winning trade, thus far. That's the chipmakers, the optical interconnectors, the turbine shippers, the liquid coolers, and everything in between."
SaaS in the Prove-It Era
Fears of a software collapse have been overblown, according to a16z, though the recalibration is real. "SaaSPocalypse? More Like SaaS-Prove-It." The SaaS category is being stress-tested in the AI era, with high-growth software companies now rare and profitability the new baseline: roughly 75% of public software companies are now profitable. "What's transpired, however, is better described as 'prove it,' than -pocalypse." Cybersecurity and observability have emerged as the runaway software winners in 2026, with the broader software market awaiting proof that AI investment converts into measurable productivity. Despite 69% of S&P 500 companies pointing to live AI deployments, only 2% disclose a tracked metric. The deployment era is early.
Private Markets: Value Accruing Off-Exchange
The private market picture is equally striking. The five largest private companies now collectively exceed $2.2 trillion in valuation, greater than the cumulative value of a decade of tech IPOs. Active U.S. unicorns have surpassed the Russell 2000 in aggregate valuation ($5.34 trillion versus $3.5 trillion). The "private-for-longer" phenomenon continues to shift value creation off public exchanges, with IPO exit value in 2026 year-to-date reaching $2.188 trillion, a new high watermark. VC dispersion has never been wider, with top-decile funds delivering net IRRs of 25% to 40.5%, while median and bottom-decile returns remain flat or negative. The power law abides: the top 1% of exits account for 84% of all exit value.
5 Key Takeaways for Advisors and Investors
- The earnings case for tech is intact. Multiples are compressed relative to history while earnings growth is accelerating. Tech is not priced for perfection; it is priced for durability with room to expand.
- AI infrastructure is a multi-year capex cycle, not a one-time event. Hyperscaler spending surpassing $1 trillion annually from 2027 means sustained demand for semiconductors, power, cooling, networking, and data center real estate. The supply chain is the trade.
- Jevons Paradox means cheaper AI accelerates adoption, not substitutes for it. As token costs fall and agentic workflows proliferate, total AI consumption is expanding faster than prices are falling. This is structural demand growth, not saturation.
- Private markets increasingly capture the compounding phase of growth. With private company valuations exceeding small-cap public benchmarks, portfolios that lack private market exposure are systematically missing where value is being created.
- SaaS requires patience, not abandonment. The "prove it" era will separate durable businesses from momentum stories. Cybersecurity and observability are already delivering. Broader software re-rating depends on AI-driven productivity metrics becoming observable and reportable.
Footnote:
1 Andreessen Horowitz Growth. "State of Markets, September 2026." a16z Growth, September 2026, https://d1lamhf6l6yk6d.cloudfront.net/uploads/2026/10/State-of-Markets-Sep-2026.pdf.
