Unlock Editor’s Digest for free
FT editor Roula Khalaf selects her favorite stories in this weekly newsletter.
Microsoft CEO Satya Nadella expressed a common opinion in the tech industry when he said Recently, large language models, the drivers behind the rise of generative AI, are becoming “more of a commodity.”
With a handful of leading model builders vying for bragging rights with each new iteration of their AI, it’s becoming difficult to separate OpenAI’s latest GPT from Anthropic’s Claude or Google’s Gemini.
That makes it even more notable that Nadella’s Microsoft just lined up behind OpenAI. latest round of financing, increasing its valuation to $150 billion. Will this moment be considered the peak of generative AI mania?
Valuing any fast-growing technology company in a new market is notoriously difficult. But the extent to which the generation AI has transformed the technology landscape and the speed of OpenAI’s emergence has left investors searching for historical benchmarks and comparisons.
First, consider what you’ve built. ChatGPT, launched almost two years ago, became a successful consumer brand almost overnight and now has 250 million weekly users. The $20 monthly subscription fee paid by a small minority has boosted their annualized revenue to $3.6 billion.
OpenAI could also be on its way to becoming a broader technology platform. Many other companies have integrated their AI into their own products and services. The tools it is creating to make its technology more useful in the business world have given it a rare opportunity in the enterprise market.
It’s tempting to draw parallels with previous startups, like Google. When the search company’s stock market value first hit $150 billion in 2006, it wasn’t the clear search winner it became, with less than half the market. Its $10 billion in revenue that year was similar to the $11 billion OpenAI projects for next year.
But this is where comparisons break down and the magnitude of the challenge OpenAI faces becomes more apparent. Google was already producing cash in 2006. OpenAI, without a functional business model, is on track to burn through more than $5 billion in cash this year, with little prospect of stopping the flow anytime soon.
Coupled with the increasing expense of training larger and larger models, the considerable computing power required to respond to user input will continue to weigh heavily on margins as it grows. It also doesn’t seem to be able to use pricing as a weapon. Although it has lowered prices rapidly to achieve greater efficiency in responding to queries, the costs of consulting other LLMs that are available through major cloud services have fallen almost in parallel.
That points to OpenAI’s biggest challenge: the lack of deep moats around its business and the intense competition it faces.
On the consumer side, Meta said last week that 500 million people now watch its Meta.AI at least once a month, a sign of the vast captive markets available to OpenAI’s Big Tech rivals. Google and Meta also have ready-made advertising businesses, which have proven to be the best route to monetizing digital audiences at scale.
ChatGPT may aim for a prime position on the iPhone, thanks to a deal with Apple. But Apple is only making the chatbot available through its Siri assistant, and even then only to handle questions that are beyond the current capabilities of its own AI models, which is hardly a recipe for long-term success. term as OpenAI tries to cement its first benefits for consumers.
Competition in the business arena is also growing rapidly. Microsoft, a close ally, is diversifying away from its initial reliance on OpenAI, while the capabilities of open source AI models have advanced rapidly, making them viable alternatives. Meta’s Llama has not yet become “the Linux of AI,” as Mark Zuckerberg suggested last week, but the risk of commoditization that Nadella warned about looms large.
At this point, it is worth remembering that generative AI is still in its infancy, and that the vast resources being invested in this technology could still hold big surprises and bring major unforeseen disruptions.
The latest OpenAI models hint at the potential. Its voice-powered GPT-4o is credited with breaking new ground in naturalistic voice interaction, potentially opening new consumer markets to AI. And he claims that his GPT-o1 is the first model capable of analyzing a complex problem and finding a solution through reasoning. This could point to a future where AI models themselves take on more of the work in an enterprise application, absorbing value from traditional software as they become more central to work life.
It’s impossible to say to what extent capabilities like these will advance and whether OpenAI will be able to maintain a significant lead in model building. But with the most powerful tech companies closing rapidly, investors backing the $150 billion group will need a strong stomach.
richard.waters@ft.com