The Next Trillion-Dollar Company Won't Sell Software — It Will Sell the Work

Silicon Valley's most influential venture firm is betting that AI will devour the $10 trillion global services economy, not just the software market

In the corridors of Sand Hill Road, a provocative thesis is reshaping how the world's most powerful venture capitalists think about the future of artificial intelligence. Sequoia Capital — the firm that backed Apple, Google, and Nvidia — is making its boldest prediction yet: the next trillion-dollar company won't be a software vendor. It will be a services firm powered by AI [1].

The claim, articulated across a series of essays and conference keynotes by Sequoia partners throughout 2025 and into 2026, represents a fundamental reframing of how Silicon Valley understands AI's commercial potential. Rather than viewing AI as the next generation of software tools, Sequoia argues that AI agents will consume the vastly larger market for professional services — accounting, legal work, healthcare billing, IT management, and consulting — replacing human labor with autonomous "autopilots" that deliver outcomes, not interfaces [2].

It's a vision with staggering financial implications. And it arrives at a moment when the AI industry faces intensifying scrutiny over whether its sky-high valuations can be justified by actual revenue.

The $600 Billion Question Meets the $10 Trillion Answer

To understand Sequoia's thesis, it helps to start with the firm's own skepticism. In mid-2024, Sequoia partner David Cahn published a widely circulated analysis titled "AI's $600B Question," warning that the AI industry needed to generate roughly $600 billion in annual revenue to justify the infrastructure being built to support it — and that actual revenue was nowhere close [3].

The gap between capital expenditure and realized returns remains enormous. Big tech companies are projected to spend a record-breaking $650 billion on AI infrastructure in 2026 alone. Yet even with optimistic revenue projections — assuming Google, Microsoft, Apple, and Meta each generate $10 billion annually from AI products — Cahn calculated a $500 billion annual shortfall [3].

But Sequoia's partners believe they've found where the missing revenue will come from: services.

"For every dollar spent on software, six dollars are spent on services," wrote Sequoia partner Julien Bek in the firm's landmark essay "Services: The New Software" [2]. The total addressable market isn't the $500 billion global software market. It's the multi-trillion-dollar services economy — a target at least an order of magnitude larger than the cloud transition that birthed the previous generation of tech giants [4].

Source: Sequoia Capital / 'Services: The New Software'

Data as of Mar 10, 2026CSV

The numbers are eye-catching. Sequoia's analysis identifies more than $1 trillion in addressable markets across just a handful of professional services sectors: IT managed services ($100B+), supply chain and procurement ($200B+), recruitment and staffing ($200B+), management consulting ($300–400B), and insurance brokerage ($140–200B). Narrower verticals like accounting ($50–80B outsourced in the U.S. alone), healthcare revenue cycle ($50–80B), and legal transactional work ($20–25B) add hundreds of billions more [2].

Copilots vs. Autopilots: The Framework Behind the Bet

At the heart of Sequoia's thesis is a distinction between two models of AI deployment that the firm calls "copilots" and "autopilots" [2].

Copilots are AI tools placed in the hands of professionals — think of a lawyer using an AI assistant to draft contracts, or a coder using GitHub Copilot. The professional remains the customer, the AI makes them more productive, and the company charges a software subscription. This is the model that has dominated AI applications to date.

Autopilots are something different entirely. Instead of selling a tool to a professional, an autopilot company sells the completed work to the end customer. The AI doesn't assist the accountant — it is the accountant. The company charges for outcomes, not seats [2].

The economic implications are profound. As Bek writes: "If you sell the tool, you're in a race against the model. But if you sell the work, every improvement in the model makes your service faster, cheaper, and harder to compete with" [2].

This insight explains Sequoia's investment allocation. While the firm has deployed approximately $150 million into foundation models like OpenAI, Safe Superintelligence, and xAI, it has poured "an order of magnitude more dollars" — over $1.5 billion — into application-layer companies, many of them services-oriented [5]. Portfolio companies in this category include Harvey (legal AI), Glean (enterprise search), and a growing roster of vertical autopilots like Rillet (accounting), Anterior (healthcare), and Mercor (recruiting) [2][5].

The Agent Economy Takes Shape

At Sequoia's AI Ascent 2025 conference, partner Konstantine Buhler outlined what he sees as the next evolution: a full "agent economy" in which AI agents don't just complete individual tasks but engage in economic transactions, communicate with each other, and develop persistent identities [4].

The vision extends beyond simple automation. Buhler envisions trillion-dollar markets created by 2030 through autonomous agents handling back-office operations, research, and creative work. "Will we see our first trillion-dollar applied enterprise software company?" he asked the conference audience. "I think the answer is yes" [1].

Three technical challenges stand between the current state and this vision: persistent agent identity, seamless communication protocols between agents, and robust security and trust mechanisms [4]. But the economic incentive to solve these problems is enormous, because the shift from copilots to autopilots fundamentally changes how AI companies are valued.

When a company sells software, it captures the "tool budget" — typically a fraction of what organizations spend on the professionals who use those tools. When a company sells work, it captures the "work budget" — the full cost of human labor in that function. The work budget dwarfs the tool budget in every professional category [2].

The Revenue Rocket Ships

The revenue trajectories of leading AI companies suggest the market is moving fast — perhaps fast enough to begin closing Sequoia's own $600 billion gap.

OpenAI is expected to generate approximately $13 billion in revenue in 2025, with projections of 2.2x growth in 2026, putting it on a path toward $20 billion or more [6]. Anthropic's growth has been even more dramatic: starting 2025 at a $1 billion run rate, the company hit $5 billion by August and $7 billion by October, with internal projections targeting $20–26 billion in 2026 [7]. Analysis from Epoch AI suggests Anthropic could surpass OpenAI in annualized revenue by mid-2026 if current growth rates hold [6].

The broader AI startup ecosystem is scaling in parallel. Cahn notes that nearly a dozen startups are on track to surpass $100 million in revenue, and that 2026 will see the emergence of what he calls the "$0 to $1B club" — companies that sprint from zero to $1 billion in revenue in record time [8].

Source: CNBC / Company Disclosures

Data as of Mar 10, 2026CSV

The private market valuations reflect this momentum. OpenAI reached a $500 billion valuation in October 2025 and has been seeking to raise additional capital at $750–830 billion [9]. Anthropic surged to $350 billion following its $15 billion Series G in November 2025 [9]. xAI, Elon Musk's AI venture, closed a $15 billion round in December 2025 at a $230 billion pre-money valuation [9]. AI startups collectively raised a record $238 billion in total funding during 2025, representing 47% of all venture capital activity [9].

The Structural Tailwinds

Several structural factors support Sequoia's services thesis. In accounting, the United States has lost roughly 340,000 accountants over the past five years while demand has grown, and 75% of CPAs are nearing retirement [2]. This labor shortage is pushing firms to accept AI faster than almost any other profession. Basis, an AI accounting startup backed by Sequoia, hit a $1.15 billion valuation and counts 30% of the top 25 U.S. accounting firms as customers [10].

In healthcare, the revenue cycle management market — worth $50–80 billion in outsourced spending in the U.S. — is "almost pure intelligence," as Sequoia describes it. Medical coding translates clinical notes into approximately 70,000 standardized ICD-10 codes with complex but definable rules — precisely the kind of high-volume, rule-governed work that AI autopilots can do at dramatically lower cost [2].

Legal services tell a similar story: 79% of legal professionals now incorporate AI tools, up from nearly zero just two years ago [10]. The $20–25 billion legal transactional market is being reshaped by AI agents handling contract review, eDiscovery, regulatory research, and case analysis.

The global AI market overall was valued at $390.91 billion in 2025 and is projected to reach $539.45 billion in 2026, with longer-term forecasts pointing toward $3.5 trillion by 2033 at a 30.6% compound annual growth rate [11]. The AI-as-a-service segment alone is expected to grow from $20.26 billion in 2025 to $91.20 billion by 2030 [12].

The Skeptics' Case

Not everyone is convinced. The first quarter of 2026 has ushered in what some analysts call a "Year of Proof," where the speculative fervor of 2024 and 2025 is being replaced by demands for return on investment [13].

A National Bureau of Economic Research study published in February 2026 found that 90% of firms reported no measurable impact of AI on workplace productivity, even as executives projected AI would increase productivity by 1.4% [14]. An MIT-affiliated report from August 2025 stated that "despite $30–40 billion in enterprise investment into GenAI, 95% of organizations are getting zero return" [14].

The Bank of England has warned of growing risks of a global market correction due to possible overvaluation of leading AI firms, noting that OpenAI more than tripled its valuation from $157 billion to $500 billion in a single year [14]. Skeptics also point to circular financing dynamics where Nvidia invests equity in AI startups that then purchase massive quantities of Nvidia GPUs — potentially inflating demand metrics [14].

Economist Owen Lamont, analyzing the AI market through his "four horsemen" bubble framework, concludes that three of four bubble indicators — overvaluation, frothy sentiment, and retail investor inflows — are present. The missing indicator: a surge in equity issuance, with U.S. companies still executing approximately $1 trillion in stock buybacks rather than rushing to sell stock [15]. However, Lamont warns that 2026 could change this calculus, with OpenAI reportedly planning a Q4 2026 IPO that could value it at $1 trillion [15].

Investor Rob May has argued directly against Sequoia's application-layer thesis, noting that application companies face margin compression from inference costs, competition from foundation model feature expansion, and uncertainty about defensibility as models improve [5].

The Tale of Two AIs

Sequoia's own Cahn offers perhaps the most nuanced perspective. In his December 2025 essay "AI in 2026: A Tale of Two AIs," he predicts a year of simultaneous delay and acceleration [8]. Data center buildouts will face construction bottlenecks — TSMC has ramped revenues 50% since 2022 but only increased capital expenditure 10% — and the AGI timeline has shifted from "2027" to "the 2030s at earliest" [8].

Yet at the application layer, adoption will accelerate. Coding assistants and ChatGPT are each expected to approach or cross $10 billion in revenue in 2026. Top AI startups are already earning over $1 million in revenue per employee — a metric that would have been unthinkable in the software era [8].

This duality — infrastructure delays coexisting with application breakthroughs — may ultimately vindicate Sequoia's services bet. If the foundation model layer commoditizes and infrastructure buildout slows, the value will migrate upward to the companies that sit closest to the customer and deliver measurable outcomes.

What's at Stake

The stakes of Sequoia's prediction extend well beyond venture capital returns. If the firm is right, the implications ripple across the global labor market, the structure of professional services industries, and the distribution of economic power.

A services-first AI economy would mean that the traditional barriers protecting professional services firms — licensing, training, relationships, institutional knowledge — could be eroded by companies that deliver the same work at a fraction of the cost. The six-to-one ratio of services spending to software spending would become a vast reservoir of profit for AI companies to tap.

It would also mean that the current public market leaders — the ten companies in the trillion-dollar market cap club, all of them infrastructure or platform plays — could be joined or challenged by a new class of companies that look nothing like traditional tech firms. They would bill like services firms, deploy like software companies, and scale like neither.

Whether this vision materializes or joins the long list of Silicon Valley overpromises depends on a deceptively simple question: can AI actually do the work? Not assist with it. Not augment it. Do it.

Sequoia is betting trillions that the answer is yes.