Antitrust Regulators Are Chasing Ghosts While Big Tech Wins the War

Antitrust Regulators Are Chasing Ghosts While Big Tech Wins the War

Washington loves a good ghost story.

The Federal Trade Commission and the Department of Justice look at OpenAI and Anthropic, see towering valuations, multi-billion-dollar cloud commitments, and exclusive distribution deals, and immediately reach for the antitrust playbook of the 20th century. They think they are staring down Standard Oil or Microsoft circa 1999. They are investigating data-sharing agreements, compute monopolies, and talent lock-ins as if these startups are holding consumers hostage by controlling the physical pipes of the internet.

It is a comforting illusion. It makes for great headlines, gives bureaucrats something to do during congressional hearings, and completely misses the structural reality of how modern artificial intelligence actually functions.

The lazy consensus in every major media outlet right now is that antitrust scrutiny will clip the wings of the foundation model labs, level the playing field for open-source alternatives, and restore fair competition to the tech sector.

That theory is wrong. It is worse than wrong; it is a distraction that protects the real monopolies while penalizing the wrong actors.

I have watched companies burn millions on compliance audits, legal maneuvering, and regulatory theater while missing the quiet structural shifts happening right beneath their feet. Regulators are obsessing over who owns the model weights today, completely blind to the fact that model weights are becoming a commodity so cheap they will soon be given away in cereal boxes.

The Fallacy of the Compute Bottleneck

Let us look at the foundational premise of the current regulatory panic: the idea that control over massive compute clusters and proprietary training data creates an insurmountable moat.

The antitrust watchdogs point to the billions flowing from cloud giants like Microsoft, Amazon, and Google into AI labs as evidence of anti-competitive vertical integration. The argument goes that by locking up the chips, these companies are building a tollbooth on the highway of human thought.

This logic collapses under a basic examination of hardware economics and algorithmic efficiency.

Compute is not oil. You cannot drill a well and corner the supply of silicon forever. Moore's Law may be limping, but algorithmic efficiency is sprinting. Every six months, researchers find ways to train models with a fraction of the compute required the year before. Quantization techniques, mixture-of-experts architectures, and synthetic data generation have democratized training to a degree that makes traditional capital moats look fragile.

When OpenAI or Anthropic sign a massive cloud deal, regulators view it as an exclusive club locking out competitors. In reality, it is a desperate race to amortize capital expenditures before hardware depreciation turns those expensive clusters into electronic scrap metal.

The real power in artificial intelligence does not reside in who owns the H100 chips. It resides in who owns the distribution choke points and the workflow integrations. And on that front, the antitrust investigations are completely looking the other way.

Why Open Source is Eating the Watchdogs Lunch

While regulators subpoena emails and analyze board seats, open-source models are quietly dismantling the commercial viability of proprietary foundation models at the margin.

Meta, Mistral, and a global army of decentralized researchers are releasing models that match or exceed commercial flagships from twelve months prior, running locally on consumer hardware for pennies. If you are an enterprise buyer, why would you mortgage your data privacy to a closed-source provider facing antitrust dissolution when you can fine-tune an open-source model on your own secure server?

The regulatory focus on Anthropic and OpenAI assumes these companies hold long-term pricing power. They do not. Foundation models are racing toward a zero-margin commodity trap faster than cloud storage did a decade ago.

When a product's marginal cost of reproduction approaches zero, antitrust law becomes an anachronism. You cannot monopolize a capability that anyone can download from Hugging Face and run on a MacBook.

Yet, the FTC and DOJ continue to swing their regulatory hammers at companies whose primary vulnerability is not market dominance, but the crushing burn rate required to stay one step ahead of open-source velocity. By forcing these labs to spend millions on legal defense and compliance overhead, regulators are inadvertently helping entrenched legacy software giants—companies that actually control closed enterprise workflows—swallow up the market without building anything innovative themselves.

The Real Monopoly is Workflow, Not Intelligence

If you want to find where anticompetitive behavior is actually locking out consumers, stop looking at the companies training neural networks. Start looking at the companies controlling the software interfaces where work actually happens.

Intelligence without distribution is just a very expensive math equation. OpenAI and Anthropic know this, which is why they are scrambling to build consumer apps and enterprise plugins. But they are playing catch-up to platforms that already own the operating systems, the productivity suites, and the cloud storage layers.

When a legacy tech behemoth bundles AI features into an existing monopoly bundle, regulators yawn. When an independent lab signs a commercial partnership to survive the capital-intensive phase of AI research, Washington screams monopoly.

It is a profound inversion of reality. The startups are being punished for trying to build sustainable businesses in a capital-heavy discipline, while the true gatekeepers of digital commerce use regulatory anxiety to entrench their suites.

The antitrust obsession with artificial intelligence labs is a symptom of a deeper intellectual bankruptcy in regulatory circles. Policymakers do not understand the technology, so they default to regulating scale. They see big numbers, big funding rounds, and big ambitions, and they assume harm.

In doing so, they are protecting incumbents, penalizing agility, and missing the actual shift in how software value is created.

The companies currently under investigation are not invincible titans crushing competition. They are high-burn research labs running a grueling marathon on a track that shifts beneath their feet every single month. By the time the antitrust regulators finish writing their comprehensive reports, the market will have evolved past the entire premise of their investigation.

Stop worrying about who owns the model weights. Start worrying about the bureaucrats who think they can legislate the pace of technological evolution.

MG

Mason Green

Drawing on years of industry experience, Mason Green provides thoughtful commentary and well-sourced reporting on the issues that shape our world.