reach the AGI (General Artificial Intelligence) is one of the great goals of Big Tech at the moment. Although expectations have been lowered regarding this concept of governing AI that does everything, Alphabet, Google’s parent company, does not want to be left behind in this race, and therefore has made the decision to reserve the computing power of its TPUs (Tensor Processing Units) to achieve this goal.
The clasee is in a term medio. For years, the cloud business has been measured by the ability to sell infrastructure to third parties. Now, with the AI race fueling demand for specialized chips, Google has opted to serve itself first, according to they count from The Register. And this is a sign of the extent to which Big Tech considers AGI a strategic objective and not just a marketing label.
In detail. The confirmation came during the presentation of results of Alphabet’s second quarter, according to the media. Goldman Sachs analyst Eric Sheridan I asked directly how the company distributed its TPUs among the customers who want to buy them and its own internal needs. “Our first priority is to ensure that we are allocating what we need to compete on the frontier of AGI development,” responded Sundar Pichai, CEO of Google. “That is the basis of everything we do,” he continued.
Asked again by another analyst, Mark Shmulik, of Bernstein, about how computing capacity is distributed between search engine, cloud and model training, Pichai insisted in the same idea of prioritizing this goal and, then, distributing it between Search, YouTube, its corporate cloud (Vertex), and other areas such as data analysis or cybersecurity.
The cloud business. That Google reserves chips for itself does not mean that the business of selling cloud is limping. Quite the opposite. Google Cloud billed $24.75 billion in the quarter, 82% more than a year before, with a profit of $8.8 billion (a growth of 214%). The company attributes this push to the strong demand for infrastructure and AI solutions, and still has a pending order book of $514 billion, that is, contracts already signed that its clients have not yet consumed.
Nor has the search engine been harmed by the emergence of generative AI, as some analysts feared. Search revenue has grown by 17% and YouTube advertising revenue has grown by 13%. Pichai defended that he AI Modede Search is, in fact, generating “more searches”, and thanks to technical improvements the company has managed to “reduce the cost of each response generated with AI” to its lowest level since that tool was launched.
And now what. Google’s appetite for chips is such that its own is not enough. And the financial director, Anat Ashkenazi, explained that the company will raise its investment forecast in AI infrastructure for this year to a range of between 195,000 and 205,000 million dollars, above the previous range of 180,000 to 190,000 million. The reason is what Ashkenazi herself described as a “supply-constrained environment,” which will force Google to turn to third-party capacity during the third quarter as a bridging solution while it expands its own infrastructure.
Pichai justified this additional expense as a long-term bet to not lose its large cloud clients, even if it means assuming specific losses for a few months in exchange for multi-year contracts that are very profitable in the long run.
However, not even these record figures have prevented a fairly significant setback. And Alphabet’s free cash flow has closed the quarter in negative, -5.9 billion dollars, something that has not happened since 2004. As expected, investors have not received this news with enthusiasm, so the company’s shares fell 4% at the close of the market.
GIPHY App Key not set. Please check settings