Might was an unusually massive month for the AI trade.
Right here’s a few of what occurred:
1. Anthropic mentioned it’s
2. OpenAI reportedly
3. Anthropic
4. Google Deepmind
Zoom out a bit, and there are a few methods to learn this. The primary is that we’ve reached the start of the tip of the AI trade’s venture-capital period and are coming into its public-markets period. That’s thrilling in the event you’re an early investor, however for now, not so significant for everybody else.
One other learn is that there’s consolidation going down, however it’s buried beneath deal buildings which might be designed to look very a lot not like consolidation. The 2 paths will converge in the event you maintain studying; let’s go down the second.
To make use of an analogy that I’m certain you’ve heard earlier than, Stainless makes the plumbing that connects fashions with all the pieces else. Which means Anthropic folded within the infrastructure layer that builders use to construct on high of what it gives; it didn’t purchase a particular product suite.
The licensing deal between Google Deepmind and Contextual AI sidesteps a evaluation by antitrust regulators because it isn’t an outright acquisition. This isn’t the primary time Google has used this construction. In 2025, it employed key employees from AI code technology supplier Windsurf and used the startup’s know-how, paying $2.4 billion in license charges. A yr prior, it arrange an analogous association with Character.AI. Like these circumstances, the transfer with Contextual AI is extra about buying experience than property.
In the entire offers, the foremost labs are after capabilities that type or strengthen moats for them, reasonably than income or customers. This sequence of unrelated offers knits collectively a compelling reply to the query of what constitutes a moat when AI is the most popular enterprise to be in, and when its proliferation is just going to change into wider.
Put it this fashion: in the event you’re in a sector the place the aggressive window is measured in months and even weeks, a buy to the tune of tens of tens of millions of {dollars} for what you’ll in any other case take two years to construct is the rational transfer, significantly when your valuation is
This issues essentially the most to individuals who, not way back, fashioned the assumption that it could be completely different to construct within the AI trade. The venture-capital mannequin assumes that the frontier is contestable, and that an agile group with a good suggestion, strong execution, and enough compute can problem incumbents which might be, properly, not as agile and even resistant to alter. However these offers sign that the AI trade’s personal incumbents are closing off the paths that challengers can use. You’ll be able to nonetheless construct on high of those platforms, however they’re making it tougher to construct round them.
For the previous three to 4 years, the enterprise market bankrolled the experiments that the AI trade’s incumbents have been both too gradual or too cautious to run themselves. (That was an odd sentence to write down, given how we’ve repeatedly described the sector as shifting at breakneck speeds.) With this association, the labs benefitted from a constellation of builders stress-testing their platforms, whereas builders obtained entry to frontier know-how and a shot at constructing one thing significant. The incumbents which might be nonetheless non-public, though not for for much longer, don’t want that setup anymore, and at the moment are able to cease outsourcing the dear components of the trade and begin proudly owning them.
The result’s a bifurcation that will trouble many: some AI builders are constructing merchandise that most likely have low defensibility, whereas others are constructing to be acqui-hired sooner or later. For the 1000’s of founders who’re mid-build, the map has modified. They should thread a needle within the gaps that main labs can’t be bothered to fill on their very own. It’s a special sort of map, one which isn’t on the frontier.
This week on the Zero Shot podcast
Hello! This is Vidhatri, the producer of Zero Shot.
For so long as I can keep in mind, one model of success round me primarily seemed like this: be part of an engineering school, get recruited by an IT-services firm, get skilled, work for purchasers overseas, possibly even go to the US for certainly one of these initiatives, and calm down.
And for years, this path paid off, altering the lives of many households. However with AI existentially cracking this mannequin, all eyes are on how IT-services corporations will reinvent themselves.
The current information about HCLTech’s $150 million wager on Sarvam is one reply to that main query. As per reviews, no VC led this spherical. However HCLTech, certainly one of India’s largest IT-services companies, is seeking to write the largest verify in Sarvam’s $300 million elevate, which valued the corporate at $1.5 billion. This can be a 7X markup on Sarvam’s final fundraise from December 2023.
However what’s HCL shopping for and what’s Sarvam promoting? What does this sign concerning the IT-services trade?
On this week’s episode of Zero Shot, host Rohin Dharmakumar debated these questions with Abhishek Pathak, the lead IT companies and web analyst from Motilal Oswal, and Kashyap Kompella, the founding father of RPA2AI Analysis, an advisory agency targeted on AI governance, dangers, due diligence, and coverage.
Tune in! You’ll be able to take heed to the episode on Spotify, Apple Podcasts, or The Ken‘s app.