The Death of the Independent Startup

On the dangers of a world of Big AI monopolies

What a quaint world we used to live in. In 1994, a 30-year-old man from Albuquerque launched an online bookstore with a series of $50,000 investments from friends and family. In January 2004, two Harvard students invested $1,000 each in a new social media website, with one student writing the code from his dorm room. During the 2008 Democratic National Convention, three former schoolmates rented out a spare air mattress in their living room, raising $30,000 of seed money by selling boxes of cereal at the convention.

In this simpler time, tech empires now worth billions could be built from the ground up with little more than a white guy, a bit of cash and an idea whose time had come. Compare this to what is emerging in the era of AI: Microsoft invested $13 billion in OpenAI; Alphabet invested $2 billion in Anthropic, then Amazon invested $4 billion into the same startup; Microsoft committed $1.3 billion to Inflection, while Nvidia has invested in Adept, Cohere, Inflection, Mistral and Runway.

Big Tech companies have replaced VCs as the principal funders of AI startups, particularly foundation model companies. In 2023, Microsoft, Google and Amazon accounted for two thirds of the $27 billion raised by AI startups. With rising interest rates and a market slowdown constraining traditional VC investors, Big Tech firms find themselves ideally placed with both the cash and the computing power to partner with emerging players. In exchange, Big Tech firms receive intellectual property and the ability to deploy the models in their existing products, also locking AI startups into using their infrastructure for years to come.

Details of the deals remain confidential, but such significant levels of investment would provide an implicit degree of oversight and control over the startups’ activities. For example, Microsoft has a 49% minority ownership in OpenAI and a non-controlling seat on the board. There is little doubt it would be setting research priorities and making strategic decisions for the firm. Further, only a fraction of Microsoft’s investment in OpenAI is in the startup’s bank, with large portions of the funding provided through cloud compute resources. Microsoft is also said to have played a key role in orchestrating the return of Sam Altman as CEO.

In early 2024, the Federal Trade Commission launched an inquiry into these partnerships, asking the companies to explain the strategic rationale for the investments and to disclose the extent of their influence over the startups. The UK competition watchdog, the Competition and Markets Authority is also closely watching deals by Amazon and Microsoft and looking into how they might impact competition in the UK.

What is an AI startup to do? Training foundation models is expensive. Sam Altman put the cost at “more than” $100 million, while Anthropic CEO Dario Amodei believes that models costing $1 billion will appear in a matter of months. Research institute, Epoch AI has estimated that on current trends hardware costs of training models could reach a staggering $140 billion by as early as 2030.

And it’s not just the money. Big Tech companies can provide the cloud infrastructure and access to compute that traditional VC firms can’t match. This creates a new form of dependency on hardware that was not as pronounced in the platform era. Big Tech is already reaping the benefits: revenue for Microsoft’s Azure was up 29% in the first quarter of 2024, while Google’s cloud platform was also up 28%, with the combined market value of these two tech giants rising more than $250 billion. Previously it was “network effects” and a strong customer base that secured platforms’ market dominance. Now their position is crystalised into the very infrastructural layer of new technology.

In the next decade, Big AI conglomerates will become some of the most wealthy and powerful organisations on the planet, perhaps even surpassing almost all nation states. Everything we feared about Big Tech monopolies pre-pandemic is now happening on steroids with artificial intelligence. While many commentators argued that AI would shake up the tech sector, it appears to be doing the exact opposite.

Market concentration is now occurring at multiple layers of the AI technology stack including chip manufacturing, cloud computing, model development and applications. Big Tech companies are moving further down the stack, investing heavily in data centres and undersea cables, while chip manufacturers like Nvidia are moving up, with their own software and cloud computing services. Such vertical integration will make forms of self-preferencing inevitable.

There are obvious negative implications for the price and quality of products in a monopolistic market. But perhaps more important still is the political impact of companies that will become too powerful to regulate effectively. A prime example is how AI startups undermined the creation of the EU’s AI Act, which became a toothless tiger on account of the 78 percent of meetings high-level EU officials held with corporate interests. Corporate Europe Observatory reported that lobbying by French startup Mistral AI and German startup Aleph Alpha was instrumental in creating an Act the startups claimed was “perfectly manageable for us” and for which “a lot of work led to significant improvements on the finish line.”

International relations scholar Swati Srivastava writes about a “sovereign awakening” of the English East India Company as it exercised an increasing level of de facto power over foreign territories during the early modern period. Criticised by Edmund Burke as a “state disguised as a merchant,” by 1800 the “company-state” of the East India Company effectively ruled one fifth of the world. What happens when Microsoft and OpenAI own a $100 billion AI supercomputer and direct $7 trillion invested in semiconductors and AI? What if the ownership and control over the digital infrastructure of everyday life is already triggering just such a sovereign awakening in the largest tech companies?

Without new competitors threatening these companies from below and political regulators holding them to account through stringent laws that constrain their power, the AI era could usher in new forms of concentrated power that were unimaginable only a few decades ago. We need investment by governments into digital public infrastructure and effective tools of competition regulation to make it possible for new startups to emerge that can access computing resources and are not instantly eaten up by their competition. The chips, so to speak, are very much stacked against us, but without further intervention, Big AI risks congealing into a new form of corporate and political power that threatens our freedom and democracy.