America's AI Giants Have No Moat — And China Knows It

Exclusive Interviews, Videos
Jay Martin: America’s AI Giants Have No Moat, And China Knows It
By Matt Morgan - October 20, 2026

Summary

Martin argues that America’s leading AI companies, Anthropic and OpenAI, lack competitive moats and are now lobbying Washington for regulatory protection disguised as safety measures to block cheaper Chinese competitors like Moonshot’s Kimi K3 model. He presents evidence that corporate customers routinely switch AI providers based solely on price, with American models’ market share on OpenRouter dropping from 70% to 30% in a year, while Anthropic and OpenAI burn billions in losses despite multi-billion-dollar revenues and owe nearly a trillion dollars to tech giants for computing power they cannot afford without continued investor funding. Martin contends this regulatory capture, modeled on how railroads used government rules to suppress trucking competition for 50 years, will force American businesses to pay inflated AI prices while global competitors access cheaper tools, ultimately protecting corporate valuations at the expense of economic productivity.

Top 5 Key Topics

  • AI companies lack competitive moats: Martin explains that businesses can switch AI providers in an afternoon without cost because all models use the same integration method, and companies like OpenRouter let clients toggle between hundreds of models instantly. Uber burned through its entire annual AI budget in three months, and a survey of 400 businesses found only 11% could forecast AI spending within 10% accuracy, prompting cost-conscious customers to shop for cheaper options and causing Anthropic’s market share to fall from 75% to 50% in nine months.
  • Chinese model undercuts American pricing: Moonshot released Kimi K3 on July 16th, which topped leaderboards for coding quality and was then offered as a free download, requiring only chip and electricity costs with no licensing fees. American models’ share of traffic on OpenRouter dropped from 70% a year earlier to 30% by summer, driven entirely by price competition, while OpenAI slashed prices by up to 80% and Anthropic followed with its own budget model family.
  • Trillion dollar promises at risk: Martin notes that OpenAI and Anthropic have committed roughly a trillion dollars to Microsoft, Amazon, Google and Oracle for future computing power, yet Anthropic posted an 8 billion dollar operating loss in 2025 against 4.6 billion in revenue and neither company turns a profit. If slower sales growth caused by price competition reduces investor appetite, those trillion dollar payments become doubtful, threatening the data center buildout that drove three out of every four dollars of US economic growth in early 2026 and underpins nearly half the S&P 500’s value.
  • Regulatory capture mirrors railroad history: Martin draws a parallel to 1935 when railroads, facing cheaper truck competition, lobbied Washington to require government permits for trucking and restrict routes, forcing many trucks to return empty and keeping shipping prices artificially high for 50 years until deregulation in 1980 cut costs by 25%. He reports that by late September, political groups backed by OpenAI and Anthropic had spent over 50 million dollars on the 2026 midterms to push for safety testing, national security bans and sanctions that would exclude overseas models from the US market without using the word ban.
  • Contradictory messaging to different audiences: Martin highlights that Anthropic’s 260 page prospectus devoted 80 pages to risks including warnings that advanced AI could threaten human survival, yet the company simultaneously seeks a 2 trillion dollar public valuation as a once in a generation investment opportunity. He argues this dual narrative, echoed by Elon Musk and Sam Altman, uses public fear to justify regulations that protect high prices and investor returns while American businesses pay more for AI than competitors in Europe or Asia who can access cheaper tools.

 



Posted in Exclusive Interviews, Videos
loading