The Corner Newsletter: Open-Weight Models and an OMI Report on the Bursting of the AI Bubble (October 20th, 2026)
Welcome to The Corner. In this issue, we explore how a new open-weight AI model from France’s Mistral will affect market structure. We also describe a new report by Open Markets on how the AI bubble will pop soon, and what policymakers should do.
Open-weights models are not a silver bullet against AI monopolies
George Colville
In recent months, one of the biggest debates in AI has focused on whether inexpensive “open- weight” models offered by Chinese corporations pose a threat to the products and business models of OpenAI, Anthropic, Google, and other U.S. corporations that have embraced a more expensive and centralized approach to delivering AI services.
Now Europe’s leading AI firm Mistral has joined the fray, with the release this month of a new high-performing open-weight model. The event marks a milestone of sorts by eroding the ability of the U.S. labs to market themselves as the only viable secure and uncensored option. Yet Mistral’s big advance also serves to underscore how — even if open-weight models ultimately prevail — such success will not guarantee an AI industry free from monopolies.
A growing proportion of businesses are now deciding that open-weight models offer better value than the proprietary models being released by U.S. tech firms. The newer models serve to commoditize the technology by making near-frontier-level capabilities available to anyone with the hardware to run it. This in turn strips the pricing power that OpenAI, Anthropic, and other closed providers have been counting on to justify the astronomical stock market valuations they have been chasing.
Successive releases, starting with DeepSeek’s infamous R-1 in early 2025, have consistently demonstrated that free open-weight models are able to keep pace with those trained by U.S. corporations operating at the frontier. Despite their creators’ attempts to position them as self-improving digital agents of both bad and good chaos, Anthropic’s Claude 5.5 series and OpenAI’s GPT-6 series today find themselves closely trailed by Alibaba’s Qwen 3.8, Moonshot AI’s Kimi K3, and Z.ai’s GLM-5.3.
Mistral Large 4 (Le Chonk) adds to that pressure by putting a non-Chinese and non-U.S. model with very high capabilities onto the market. While Mistral is immediately catering to the increasingly sovereignty-focused European market, it will also likely benefit as businesses from both the U.S. and other regions embrace a model from a corporation able to inspire more trust than its U.S. and Chinese rivals. Mistral’s model, for instance, addresses both legitimate concerns about bias and censorship in Chinese models, and theoretical — albeit yet unproven — concerns about possible backdoors.
Open-weight models can execute many economically valuable tasks just as well as frontier models do and are being used to do so by a growing number of major businesses. They also offer a very concrete technical advantage: because their weights are open, they can be fine-tuned by companies and governments to perform highly in the specific tasks and domains relevant to them. This means higher performance and significantly lower running costs, helping companies cut ballooning AI bills.
At the same time, open-weight models give users far greater control over where their data is hosted and processed. This issue has come to the fore after a leading mathematician accused OpenAI of training on his data, leading businesses, including Palantir, NVIDIA, and Booz Allen Hamilton, to suspend their use of proprietary AI models.
Yet it is becoming increasingly clear that the open-weight model simply shifts much of the monopoly problem to different levels of the tech stack. Even when running open-weight models, a majority of users still rely on infrastructure that has been deeply concentrated. Most businesses and public services, for instance, will continue to rely on cloud infrastructure controlled by Google, Microsoft, and Amazon. This in turn means they will connect their AI model to the databases, enterprise software tooling, and identity and security systems that they have already procured from cloud platforms within existing or amended cloud contracts.
Moreover, these same dominant players will be able to exploit their existing dominance of cloud platforms to promote their own open-weight AI products and steer business to integrate their open-weight systems with third-party proprietary products sold on these platforms. Not only is this more convenient for procurement, but it helps reduce the apparent cost of AI by folding it into renewal conversations for contracts customers are already committed to. For businesses operating in regulated sectors with significant compliance requirements, the fact that these three cloud giants have already adhered to many of these requirements also gives them an advantage, incentivizing procurement through them rather than through standalone operators such as OpenAI or Anthropic.
Open weights do improve portability, allowing customers to more easily move their business, for instance from AWS to an AI-specialized neocloud like CoreWeave or even in-house hardware in ways they can’t with a closed API. But in practice, the gravitational technical and contractual pull of the cloud giants undercut most of the theoretical impact of portability.
Ultimately, this means that even if AI models themselves become commodified, customers may still find themselves paying ever-higher prices to the cloud giants. The monopoly chokepoint has simply moved elsewhere in the chain.
Open Markets Releases Report, “Brace for Impact: Protecting People and the Economy When the AI Bubble Bursts”
The Open Markets Institute released a new report, “Brace for Impact: Protecting People and the Economy When the AI Bubble Bursts,” examining the rapid growth of the AI sector. The report by Open Markets fellow Matthew Scherer breaks down how a handful of tech companies and their financial backers are fueling the AI bubble, outlines vulnerabilities in the financial system that could intensify an AI crash into the next financial crisis, and provides steps that policymakers can take to mitigate the damage and protect everyday people from bearing the costs of a bubble-induced crash. The Americans for Financial Reform Education Fund is a co-publisher of the report.
Recent reporting highlights that AI warning signs are already emerging. OpenAI’s annualized revenue was a full $20 billion below a figure widely reported just weeks earlier, and Fitch Ratings, a U.S. credit-rating agency, warned that any sharp correction of AI stocks could plunge the U.S. into a recession. “Silicon Valley and Wall Street are making enormous bets on AI, and they’re finding increasingly risky ways to keep the money flowing,” Scherer said. “But when this bubble bursts, it won’t just be billionaire investors who suffer. Workers could lose their jobs, families could see their retirement savings evaporate, and communities could face the fallout from another financial crisis.” Read the report here and listen to the audio version of the report here.
📝 WHAT WE'VE BEEN UP TO:
Open Markets senior legal analyst Daniel Hanley and AI policy lead Arnav Rao published an article in the Washington Monthly arguing that AI companies do not need an antitrust exemption to coordinate on safety. Instead, they note that the corporations can use a government-supervised standard-setting organizations, or SSO, as the forum for such a conversation. SSOs, Hanley and Rao write, “govern everything from electrical plugs and fire safety protocols to specifications for medical device testing.” They also argued that the SSO model offers a test of the goodwill of the corporations that dominate AI, saying, “If frontier model companies refuse [to set up an SSO], it would reveal that their true interest is not coordinating on safety, but rather cementing their power over this nascent, increasingly important industry.”
Georgetown University’s Initiative on Artificial and Democratic Citizenship selected Courtney Radsch, the director of the Center for Media and Digital Governance at OMI, as one of its 2026-27 fellows focused on the question of what it means to be a democratic citizen in an age increasingly shaped by AI.
Open Markets Europe research fellow Claire Lavin published an article in Concurrences urging the European Commission to update merger guidelines, currently under review, so that they will clearly curb market power and not serve the interests of powerful corporations.
Open Markets Institute Europe director Max von Thun appeared on the Marketplace podcast to discuss delays in Europe involving new AI agents from OpenAI and Meta. Pushing against narratives that brand Europe as “anti-innovation,” Von Thun noted powerful new technologies should be tested and regulated much the same way we expect new drugs and vaccines to be vetted before they come to market. He also said Europeans receive better versions of products that collect less data and are easier to use with third-party apps and devices.
In an interview with Le Grand Continent, Open Markets Executive Director Barry Lynn argues that concentrated corporate power, particularly among major technology platforms, poses a growing threat to democratic institutions and individual freedom. Lynn discusses the revival of U.S. antimonopoly enforcement, the need to curb platform control over information flows, and opportunities for stronger transatlantic cooperation to challenge concentrated economic power.
Barry Lynn spoke at the US vs Billionaires convening in San Francisco on October 5, on a panel that focused on how to solve the extreme and growing inequality of wealth and power in the US. Lynn called on participants to focus on striking concentrations of power at their root, in libertarian legal philosophy, and said that doing so would make it much easier to address impose new taxes and to reform campaign finance laws. Other speakers included Congressman Ro Khanna, reformist billionaire Tom Steyer, and author and activist Saru Jayaraman, director of One Fair Wage.
CMDG director Courtney Radsch spoke at the Mediaspace Global Changemakers’ Summit on October 7, joining leaders from media, technology, regulation, and advertising for a discussion on how AI is increasingly controlling the flow of information. The summit focused on what policymakers, industry leaders, and innovators can do now to build a more competitive, fair, and sustainable digital market by 2030.
Dr. Radsch moderated “The Data Center Outrage: Can AI Survive Without Data Centers,” a panel discussion at the University of Baltimore examining the economic, energy, and public-interest questions surrounding data center expansion. The October 14 event brought together experts in data centers, energy policy, and AI to discuss whether the current model of AI development is sustainable.
Open Markets senior legal analyst Daniel Hanley gave a guest lecture on the history and purpose of the antitrust laws at a political economy of regulation class at Johns Hopkins University. He specifically focused on how they constrain concentrated private economic power, shape the structure of markets, and how they affect our liberty and democracy.
The Sling, a website offering analysis and commentary on antitrust issues. appointed Hanley to its editorial board.
TheStreet cited Open Markets Institute’s warning in a letter to the Federal Trade Commission that transparency alone is not enough to address surveillance pricing, as retailers gain new tools to tailor prices using consumer data.
Rep. Pramila Jayapal introduced a National AI Charter Act framework that would require AI companies to obtain a federal charter and comply with enforceable safety, privacy, competition, and public-interest obligations. CMDG director Courtney Radsch endorsed the proposal, saying, “Adopting the charter model offers a sweeping yet adaptable way to welcome innovation while imposing enforceable obligations for safety and fair competition, placing necessary checks on private power, and ensuring corporations develop AI in the public interest.”
Open Markets Institute joined more than 40 labor, consumer, civil rights, and public-interest organizations in a letter urging senators to reject federal judicial nominees the coalition says have records hostile to workers and consumers.
🔊 ANTI-MONOPOLY RISING:
Michigan’s attorney general filed an antitrust suit against Blue Cross Blue Shield of Michigan for using its position as the state’s largest insurer to drive up health costs for patients. The state’s suit argues that the insurer illegally cooperated with other Blue Cross Blue Shield companies to limit competition, leading Michigan’s state government and others to pay inflated prices for health coverage. (Wall Street Journal)
Infant formula manufacturer Perrigo has agreed to pay $10 million to resolve a class action antitrust suit alleging it conspired with Gerber to block new entrants selling store-brand infant formula. (Reuters)
📈 VITAL STAT:
$5 Billion
The value of an IPO now being scrapped by cloud-computing company Firmus Grid, which is backed by both Nvidia and Blackstone and cited unfavorable market conditions. The IPO would have been Australia’s largest initial public offering in three decades. (Wall Street Journal)
📚 WHAT WE'RE READING:
Anti-Fascist Economics: In Defense of Affordability, Dignity, and Democracy — Economist Isabella Weber proposes an economic framework to combat the rising inequality and overwhelming corporate power that threaten our democracy and fuel the rise of fascism. Hearkening back to the work of mid-20th century economists, Weber’s anti-fascist economics is focused on people’s needs, affordability, and dignity, rather than the corporate pursuit of profits.