Before AI breaks the market, let’s break the monopoly

The last time a handful of companies controlled the arteries of the economy, they were shipping oil, steel, and railcars. Today, they’re shipping compute, data, and large language models.

Seven firms (Microsoft, Google, Amazon, TSMC, Oracle, NVIDIA, and OpenAI) now own the infrastructure of the AI era. They invest in one another, sit on each other’s boards, and recycle capital in a closed loop that locks out challengers. Rockefeller’s Standard Oil owned the refineries. Today’s giants own the GPU refineries and data highways. Same logic, different century. The question isn’t whether this creates risk; it’s how much damage happens before regulators wake up.

Microsoft funds OpenAI, then bills it for compute, and Google and Amazon both bankroll Anthropic while competing for its workloads

Competition has been replaced by co-investment. Microsoft bankrolls OpenAI, then bills it for Azure compute. Google and Amazon both fund Anthropic while competing for its workloads. Nvidia sells GPUs to everyone, then invests in those same buyers.

One European bank told me it had to delay its internal AI rollout by three months because the only GPUs available were tied to a specific cloud provider’s credit package. The delay wasn’t about technology. It was about market power. Venture term sheets increasingly include “free” cloud credits that must be spent on the investor’s platform. On paper, that looks like growth. In reality, it’s a closed-loop economy where dollars never leave the system.

Google pays Apple $20 billion annually to stay the default search engine while early-stage AI founders scramble for Series B funding

Capital that could fuel new entrants is trapped at the top. Google pays Apple $20 billion a year to stay the default search engine. Meta and Alphabet spent a combined $90 billion on stock buybacks in 2024 while AI startups struggle to raise follow-on rounds.

A healthcare company I spoke with recently built its AI pipelines entirely on one vendor stack. When pricing changed mid-year, it spent six weeks rebuilding. Portability is no longer an IT convenience; it’s a survival strategy. Another software scale-up accepted a strategic investment with generous training credits. Six months later, the same investor launched a competing product on the same platform. The founder’s takeaway: Read the conflict clauses as carefully as the valuation.

AI’s trillion-dollar valuations rest on expectations, not earnings, and we’ve seen this movie end badly before

AI’s combined market cap rose more than $5 trillion in 2024. Those valuations rest on expectations, not earnings. The difference from the dot-com bust is that AI is now woven into everything. One major failure could ripple through the entire economy, pulling CFO spending plans and stalling deployment just as enterprises commit to modernization.

Real competition means firms win customers, not board seats, and challengers can rise without being absorbed

Healthy markets reward those who out-innovate, not those who out-invest. Platforms should be neutral arenas, not competitors on their own turf. Capital should fund research, skills, and affordability, not endless buybacks. Challengers must be allowed to rise without being absorbed or suffocated.

Look back at the semiconductor wars of the 1980s and 1990s. Intel, AMD, and ARM pushed each other relentlessly, and the result was decades of true innovation. Today’s AI chip race feels more like mutual insurance than rivalry.

Regulators slept through the last consolidation wave; they must act now before the AI market hardens

Regulators slept through the last wave of tech consolidation. By the time they woke up, the cloud was already cornered. This time, they can act before the foundations harden.

That means banning overlapping board seats among competitors, limiting circular cross-ownership, scrutinizing acqui-hires that quietly remove competition, and enforcing penalties that make compliance cheaper than misconduct. Most importantly, we need open standards so enterprises can switch AI providers without rewriting their businesses. We must encourage portability through open standards for model inference, data formats, and agent orchestration. Procurement should be able to switch without rebuilding the stack.

Productivity is rising, but prosperity isn’t; when citizens stop believing innovation creates opportunity, democracy becomes brittle

AI’s concentration problem mirrors a larger economic crisis. Productivity is rising, but prosperity is not. The middle class is shrinking while corporate wealth consolidates. Mobility is declining, middle-income households are being priced out of opportunity, and citizens see innovation gains accruing to a narrowing set of firms and investors. When people stop believing that markets are fair, the political system absorbs the stress. This isn’t only a technology policy issue. It’s a social stability issue.

The Bottom Line: AI should democratize intelligence, not monopolize it. The window for action is closing.

AI gives us a second chance to design capitalism that scales innovation without killing competition. The goal isn’t to punish success. It’s to keep the market open, fair, and resilient.

Regulators must act now to enforce accountability on monopolists, reward investment in people and R&D, tax capital more fairly, and rebuild a middle class that shares in technological progress. The window for preventive action is still open, but it’s closing fast. If we miss this moment, we’ll end up performing surgery on monopolies instead of building smart rules that keep markets open.