Why a Do-Nothing Congress could be great for markets | E2359

Markets rose after the election despite expectations of a 2–4% decline. DA Wallach argued that a divided, gridlocked Congress could support markets by making abrupt policy changes less likely and the business outlook more predictable. He attributed higher long-term interest rates more to uncertainty and fiscal unsustainability than to AI borrowing crowding out other activity. That predictability has limits: Wallach warned that data centers could become a bipartisan political target because concerns about jobs, electricity, and water may resonate more readily than arguments that technology creates new work.

Ilana Gold said opposition to data-center construction could turn expansion into a state-level fight and make national-scale development harder. She cited Elon Musk’s statement that the U.S. could run out of power in 2026 and argued that constraints or regulation would push startups toward alternative energy, nuclear power, and more efficient chips; restricting data centers would not eliminate demand for electricity or AI compute. An NBC exit poll reportedly found 56% of voters considered AI and data centers a crucial or major factor in their vote, while Democrats reportedly spent more than $100 million on ads linking Republicans to big tech and data centers. Another speaker countered that data centers generate substantial tax revenue, no longer seek local tax breaks, and may guarantee electricity prices will not rise.

The speakers saw AI adoption as a contest between visible productivity gains and fears about job losses, costs, and concentration of benefits. One said their company spends about $40,000 on AI tools and estimates they generate $2 million in value; a platform that might have taken four or five people two years to build was completed in three months without developers or designers. Another cited reports that Canva cut 22% of its staff that week, attributing the cuts to AI-driven compression of paid work. Alana cautioned that customers may press businesses using third-party AI inference to lower prices and cut costs. The discussion also questioned whether AI companies’ high gross margins can persist: competition and cheaper open-source models may put pressure on pricing, while practical agents that save users time or money could make AI’s benefits more tangible.

AI’s economics may shift from broad deployment to measuring cost, capacity, and output per worker. Tomash said hourly costs for state-of-the-art Blackwell 300 GPUs rose about 120% in six months, from roughly $5 to $13, partly because of the election. He forecast an AI shortage in 2027, with demand growing faster than 200% annually against about 81 gigawatts of capacity next year, roughly 60% devoted to training. He cited training runs costing $300 million to $500 million, Vera Rubin chips being twice as productive, and intelligence per watt doubling annually, alongside overall costs falling 400-fold. If supply remains constrained, AI could be sold more like bandwidth, with customers paying by question volume and response speed rather than receiving unlimited access. Local inference may also expand: Tomash cited about 10 million Olamah users and a study finding that 90% of non-coders’ AI work can run on local models.

The speakers connected technology and regulation to dealmaking and venture capital. One warned that four years of little M&A under Lena Khan had stalled acquisitions they considered sensible, and said a major deal might need to be executed by June or July next year to beat an approval process that can take 18 months; they viewed 2027 as potentially the last year before a five-year freeze. Sequoia is closing its London office, ending Arc and its Scouts program, combining early- and growth-stage groups, and has announced $10 billion in new funds. The speakers called the combined approach sensible in a market where seed rounds can range from about $5 million to $500 million, but questioned whether a $10 billion fund can deliver historically high venture returns. They also noted that some AI-built businesses reach $10,000–$100,000 in monthly revenue and may be profitable, reducing founders’ need for venture funding.

A broader concern was that companies staying private for longer deny ordinary public-market investors access to growth while intermediaries profit from secondary trades. One speaker argued for earlier public listings and financial disclosure, citing a company valued at $5 billion with $800 million in revenue, and said the decline in publicly traded companies was an unintended consequence of higher shareholder thresholds under the JOBS Act and greater access to private capital. The speakers supported wider access to startup investing but cautioned against putting 401(k) savers into private-company vehicles. They also discussed tokenized equities, debt, and synthetic markets as possible ways to broaden access and enable around-the-clock trading, while expecting substantial U.S. regulation. Highly leveraged products may attract people seeking rapid gains amid wealth inequality, but the speaker explicitly warned that such bets are unlikely to be a sound retirement strategy.