ˮ����

CHT prepared a background briefing for the Office of the Director of National Intelligence (ODNI) in April 2024. ODNI compiles a public every four years, timed to release with the incoming administration. It is meant to outline global trends impacting national security for the coming 20 years, forming a fundamental piece of background research. 

CHT’s contribution looked at how global power structures are shifting, how these shifts impact the digital ecosystem, and the implications of this for resiliency and tech governance. The paper focused primarily on the structures and incentives for technology companies. 

The following article, cross-posted from , builds upon some foundational ideas in this paper, while taking a more narrow look at power consolidation in the top AI companies. 

‍

Background

In a landmark , yesterday a federal judge declared Google a monopolist, assessing that it used its dominant position to box out competitors and charge its advertising customers higher prices. The company will appeal the ruling, and the case will likely take years longer to resolve. Meanwhile, the problem of consolidation of power in the tech industry– especially in the hands of just a few players, including Microsoft, Google, Amazon, Apple, and Meta– will continue into the era of artificial intelligence, as these companies hoard infrastructure, resources, and talent.

This dynamic poses a significant challenge to society. To understand why, consider the case of Microsoft and its relationship with OpenAI.

‍

A very good email

In the early morning hours of June 12, 2019, Microsoft chief technology officer Kevin Scott drafted an to the company's CEO, Satya Nadella, and founder Bill Gates. Under the subject line “Thoughts on OpenAI,” Scott wrote that he had previously been “highly dismissive” of OpenAI’s efforts — and that that had been a mistake.

“When they took all of the infrastructure they had built to build [natural language processing] models that we couldn’t easily replicate, I started to take things more seriously,” Scott wrote. “And as I dug in to try to understand where all of the capability gaps were between Google and us for model training, I got very, very worried.” Scott would add later in the email that some of OpenAI’s capabilities were getting “scarily good,” and described Microsoft as “multiple years behind the competition in terms of [machine learning] scale.”

Nadella replied later that day, “A very good email that explains why I want us to do this… and also why we will then ensure our infra folks execute.” Nadella copied Microsoft’s chief financial officer Amy Hood on his reply: “Amy - fyi.”

Six weeks later, Microsoft announced a $1 billion investment in OpenAI.

This email exchange is just one example of how, amid the flurry of headlines about new players in the tech industry, a deeper and exceedingly familiar pattern is at play. Power is still being consolidated in the hands of a few ultra-powerful companies — Microsoft, Google, Amazon, Apple and Meta — even as fresh-faced startups like OpenAI are propped up in the public eye, touted for their exciting visions and innovation. The age of artificial intelligence has led to increasingly subtle power consolidation tactics from the Big Five — they are hoarding infrastructure, resources and talent, all while trying to outpace antitrust enforcement, like a game of cat and mouse.

Should these behemoth companies continue to accumulate power, we face a future where industry competition is dampened, and “innovation” is defined as companies looking for new ways to monetize their accumulated resources, power and scale, instead of as entrepreneurs bringing forth novel ideas and problem-solving technologies. We will inhabit a future where anyone (or anything) not affiliated with one of the Big Five will face creeping irrelevance — including, potentially, nation-states themselves.

‍

Dominance, for sale

Subtle consolidation tactics start with investments. Amazon may not have acquired AI startup Anthropic, but it has, so far, $4 billion in the company. Microsoft doesn’t own OpenAI, but its initial investment locked OpenAI into Microsoft’s cloud computing platform, and follow-up investments of over $10 billion have led Microsoft to claim nearly 49% of OpenAI’s equity.

These massive corporations have thus found ways to reap the benefits of an acquisition — and the hype of a startup — without taking on the risk of acquiring a company. Even hardware juggernaut Apple has found a way to get a piece of the hype, with OpenAI to weave the firm’s software into its new fleet of iPhones. Now, the success of OpenAI or Anthropic feeds directly into the success of Microsoft, Apple and Amazon, thereby amplifying the power of already-monolithic corporations. Cloud computing deals only deepen said power, by allowing companies such as Microsoft and Amazon to hedge their direct investments in smaller firms, while still profiting off of these relationships. The Big Five’s influence, and infrastructure, thus become ubiquitous.

Beyond investments and partnerships, what else drives power consolidation in tech, in ways that shut smaller players out?

‍

Own the infrastructure

As seen in Microsoft’s 2019 , major companies are consolidating power through infrastructure, in the form of chips and cloud computing. In the race to develop and deploy AI systems, competitive advantage calls for raw, hard processing power in order to train powerful foundation models — and this processing power is expensive. NVIDIA’s chips are one of the most sought-after resources in tech, and tech investors have even begun plugging their ability to acquire chips as a key differentiator for themselves in the space. If large language models are the exclusive club everyone in tech wants access to, then club-goers need to be ready to fork over a pricey cover charge at the door to get in.

OpenAI CEO Sam Altman himself raised eyebrows earlier this year when he that he is seeking trillions of investment dollars in a quest to push OpenAI’s models to previously unthinkable levels of sophistication. For smaller companies in search of compute, maintaining a competitive edge in this kind of financial environment is nearly impossible. Without access to the same infrastructure and compute as these larger, Big Five-backed startups — or without a meaningful decline in the price of compute — smaller firms simply can’t compete in a meaningful way.

‍

Dominate with data

Power is also being consolidated when it comes to data, which is required to train and refine sophisticated large language models. Big Five companies such as Google and Meta are leveraging the scale of their existing businesses — including vast social media platforms (with trillions of posts), search, and ecommerce — to surface new heaps of data in order to train models such as Gemini and Llama.

But smaller companies don’t have access to those mountains of data. In their quest to stay ahead, they at times turn to a dubious yet common business practice in tech — wading into legal gray areas in order to develop their products. One example is Perplexity, a buzzy AI-powered “answer engine” that found itself in hot water this summer. that the company used bot crawlers to discreetly scrape content from websites, without the site owners’ permission, in order to produce its compelling “answer” results. (At the time of this writing, Perplexity had followed up the controversy with a change of tune, announcing aimed at sharing ad revenue with publishers. Time, Fortune and Entrepreneur are some of the first publications to sign up for the program.)

Firms such as OpenAI have been in similar hot water, accused of stealing intellectual property, private information and copyrighted material for their large language models. Copyright infringement have been filed left and right. But smaller and mid-sized startups may not have the legal manpower — or deep enough pockets — to take on lawsuits resulting from their data-scraping practices, even as larger firms like OpenAI engage in their own version of it, as well.

What’s more, in the battle between AI companies and media companies, olive branches are being extended in the form of content licensing agreements, with bigger names being favored. Reddit has an AI content licensing agreement with Google; The Atlantic and Vox Media have licensing deals with Microsoft-backed OpenAI. Smaller companies may not have the resources to buy the silence of media companies, leaving them to face fierce litigation over their data practices, without the forgiveness structure of a supportive licensing arrangement down the road.

‍

Brains equals brawn

Finally, talent remains one of the most valuable resources in Silicon Valley — and the Big Five are keen on consolidating power there, as well. Earlier this year, Microsoft Inflection co-founder Mustafa Suleyman as CEO of Microsoft AI. To many, this amounted to a soft acquisition of Inflection, given Suleyman’s importance at the firm. Similarly, Amazon the co-founders of Adept, and struck a licensing deal for the company’s software to placate Adept’s investors. Most recently, Google the co-founders of Character.ai and its researchers in an analogous deal. Even when sought-after talent in tech branches off to launch a new company, success — and the promise of deals rich with resourcing — leads them back into a revolving door between the Big Five and their investments.

‍

What’s lost when Big Tech wins?

When power is consolidated like this among behemoth tech companies, it doesn’t just impact product markets and consumer experiences. This level of power consolidation has the potential to dwarf the power of nation-states, and undermine democratic processes themselves. Should power continue to be hoarded in this way — especially in the age of AI — society faces several key consequences.

For one, competition will be dampened in the tech industry, which in turn can affect consumer welfare and the welfare of businesses and governments that may be applying AI products at scale across their institutions. While prices for AI products have started low, it’s likely that prices will increase as companies seek to recoup the costs of their expensive battle for customer acquisition — mirroring trends in rideshare, streaming, and delivery services.

Secondly, the centralization of power, and resources, can heighten our already-fragile digital ecosystem. When we rely on a handful of large companies — or just one — to power our digital lives, we risk single points of failure that ripple outward, as seen with this summer’s CrowdStrike fiasco in Windows systems, which crashed computers across the globe. Crowdstrike’s market share reˮ����ed only a fraction of computer systems — what might happen when financial firms worldwide all rely on similar AI algorithms to automate trading decisions? What might happen if that algorithm goes awry?

Finally, we risk having these powerful corporations develop powerful technology with their financial bottom line as the main incentive. This means monolithic companies prioritizing their efficiency and profits, all while the vulnerability of certain groups, the public’s well-being, and the health of our democracies are left as secondary considerations, or afterthoughts altogether.

Beyond these practical consequences, these trends in power consolidation erode away at true American entrepreneurship, one of the core elements of the American dream. The companies that are currently set up to win in tech right now aren’t the ones with the best skills, the most exciting and innovative ideas, or the deepest work ethic and vision. Instead, the companies that are set up to succeed are the ones with the greatest access to finite, expensive resources.

To use Kevin Scott’s language, these huge tech companies are getting “scarily good” at consolidating power. As they deepen the divide between the tech industry’s “haves” and “have nots,” what benefits do ordinary citizens, and nation-states, really gain? Consumers may benefit from lower pricing in the short term, and businesses may enjoy modest productivity gains from the AI systems sold to them by Microsoft, Amazon and Google.

But when we widen the lens through which we evaluate the impacts of power consolidation, we gain a better understanding of its true impact. When power in tech is structured in this way, is it truly serving society? Should things continue to trend in this direction, the Big Five will have more than ever, and society risks being left without much at all. 

This piece was originally posted in .