With Harry Stebbings, Jason Lemkin, and Rory O’Driscoll
The landscape of artificial intelligence is currently defined by a singular, gravitational force: an unprecedented flood of capital and elite talent that is forcing every sector of the global economy to pivot. In a recent deep-dive discussion, industry veterans Harry Stebbings, Jason Lemkin, and Rory O’Driscoll analyzed the current state of the AI and B2B markets, revealing a "parallel universe" where traditional business models are being dismantled in real-time. From the sovereign ambitions of the Chinese AI sector to the existential struggle of the "number three" model providers, the conversation highlighted a market stretched to its limits.
The Chronology of an Industrial Shift
The current state of AI is not merely an evolution; it is a rapid-fire sequence of events that has caught even the most seasoned observers off guard. Over the last few months, the industry has witnessed:
- The Talent Exodus: Within a 48-hour window, Google DeepMind lost two of its most critical scientific pillars—Noam Shazeer and John Jumper—to Anthropic. This move sent shockwaves through Silicon Valley, signaling that top-tier researchers are prioritizing autonomy and shipping velocity over the relative comfort of incumbent bureaucracy.
- DeepSeek’s Sovereign Round: DeepSeek secured $7.4 billion in funding at a $50 billion valuation. The structure of this deal, which excludes voting rights for private investors and concentrates power within the Chinese state, underscores a burgeoning global reality: AI has become a matter of national sovereignty, not just market share.
- The Infrastructure Crunch: DRAM prices have surged 90% to 95% in a single quarter. This is not just a line item in a tech budget; it is a ripple effect that will inevitably increase the cost of consumer hardware, from iPhones to enterprise data centers.
- The Rise of "Agentic" Workflows: As demonstrated by practical applications—such as building an autonomous "AI VP of Finance" in mere hours—the industry is moving away from simple chatbots toward complex agent systems that handle end-to-end business operations.
The $725B Question: Can the Math Hold?
The central tension of the current market lies in the disparity between capital expenditure (CapEx) and realized revenue. Goldman Sachs projects cumulative AI CapEx from 2026 to 2031 to reach $7.6 trillion. When contrasted with current total AI revenue—which remains well under $100 billion—the "math" of the industry appears increasingly speculative.
Hyperscalers are currently spending roughly $700 billion annually on infrastructure. For this to be a sustainable business, the market must generate at least $1 trillion in annual revenue to account for electricity, hardware, and operational overhead. To reach this figure, AI must displace roughly 7% to 8% of the total U.S. labor force. The question is no longer whether AI is "smart," but whether it is productive enough to trigger this level of labor displacement, or whether the current investment cycle will result in a painful correction.
The Peril of the Number Three Position
History shows that technology markets rarely support dozens of winners. They inevitably collapse into oligopolies: a leader, a runner-up, and, occasionally, a niche third player. In the current LLM market, being the "number three" closed-source provider is becoming a precarious position.
As enterprises adopt "model routing"—the practice of directing workloads to the most efficient model for a specific task—the number three player is squeezed from both sides. Number one and two models capture the high-end frontier, while open-source models, often subsidized by state-backed or venture-funded ecosystems, undercut the cost of the mid-tier. When an industry is ground down by a 5x-cheaper alternative, the market leader may see a dip in profits, but the number three player often faces total obsolescence.
Implications for B2B Founders and Investors
The rapid evolution of AI has forced a re-evaluation of what makes a company "defensible."
1. The Death of the "Margin Later" Strategy
For the past three years, founders were rewarded for hypergrowth, even at the cost of negative gross margins. That era is closing. Investors are now scrutinizing delivery costs with renewed intensity. If a company’s unit economics are not sound, or if the path to profitability is not clearly paved by efficiency rather than subsidized growth, raising subsequent rounds will become increasingly difficult.
2. The Myth of the Moat
Traditional switching costs—once considered the ultimate moat—are crumbling. When an LLM can perform complex data migrations or systems integration in days, rather than the months or years previously required by consultants like Accenture, incumbent software vendors lose their leverage. Founders who rely on "integration pain" as a protective barrier are finding themselves vulnerable to AI-native competitors that can perform the same work at a fraction of the price.
3. The "Master of Agents" Era
The most successful teams are not firing their staff; they are "agentizing" the work that humans find repetitive or unmanageable. CRM hygiene, invoicing, and contract follow-ups are being handed off to agents. The job of the future is not "prompt engineering"—a skill that is rapidly being commoditized—but "Master of Agents." This role requires the ability to troubleshoot, refine, and manage AI systems that occasionally falter, ensuring they reach their objective despite the inherent unpredictability of LLMs.
4. Intensity as a Competitive Advantage
The debate over remote work has shifted. While remote work was the norm for the post-pandemic years, many high-performing startups are now gravitating toward small, elite, in-person teams. The argument is simple: the pace of AI innovation is so relentless that a distributed team working 20 hours a week cannot compete with a high-intensity, in-person team working six or seven days a week. It is a choice between a comfortable, steady career and the high-risk, high-reward pursuit of market dominance.
Official Perspectives and Market Sentiment
Rory O’Driscoll, reflecting on the governance of the DeepSeek round, noted that the structure was entirely rational: "The only people getting voting rights are the only people who don’t need them." In this view, state-backed entities do not need shareholder votes because they exercise sovereign control.
Meanwhile, the decline of consulting giants like Accenture—down roughly 40% this year—serves as a canary in the coal mine for the services industry. AI is directly targeting the "white-collar BPO" (Business Process Outsourcing) model. As AI-native SIs (Systems Integrators) begin to undercut traditional firms by offering to do for $15 million what once cost $80 million, the margin structure of the traditional consulting industry is facing a terminal threat.
Conclusion: The Path Forward
The "AI bull market" trap is classic: one can be intellectually correct that the sector is stretched while the narrative and capital continue to push prices higher for years. The advice for founders and investors remains pragmatic: deploy into the demand, build the most efficient company possible, and strive for liquidity before the cycle reaches its inflection point.
The industry is currently engaged in an endless series of sprints. Every time a company finds a moment to breathe, a new inference chip or a breakthrough model is announced. In this environment, the only viable strategy is to build a team capable of constant adaptation, maintain a laser focus on unit economics, and accept that in the age of AI, there is no longer a "middle ground." One either embraces the intensity of the frontier or risks being left behind as the rest of the market pivots.
