Q3 2026
This is the headline from early Q3. Palantir's CEO Alex Karp put a finer point on it in his CNBC viral moment, but it has gained traction everywhere. Companies are realizing that they need to protect their IP, their edge, their data. This is being called "AI sovereignty", loosely defined as a company's ability to control its AI models, governance, data and operations. The impetus? Companies worry that leaning on the frontier labs enables the labs to learn from what organizations are building by watching their usage, and then rebuild it faster and cheaper, only to monetize it themselves. The enterprise isn't having it.
Still waiting on AI to transform your business while your data is a mess? Many companies we're talking to are taking a harder look at their data quality and governance, and realizing its not at the level it needs to be to leverage AI. Some are even going further and concluding an initiative isn't worth starting until the data story is sorted. This is not a new story, but we believe it's a sign of a maturing industry. A year ago every board wanted an AI strategy. Now they want an AI outcome, and they're learning the result lives in the data. Fix the foundation or stop pretending.
Companies are reassessing their org charts, and one of the sharpest lines cutting across is dividing people into builders vs. non-builders. Yes, it's reductive. It's also the lens leaders are using to decide who is adding value and who is overhead. It might not be in any job description or explicit org chart, but worth considering: what are you building?
In the span of about 3 months, AWS committed $1 billion to a new forward-deployed engineering unit, and Microsoft put $2.5 billion and 6,000 people behind its "Frontier Company" to embed engineers inside customers. Both follow the forward-deployed ventures OpenAI and Anthropic launched earlier this year, and the model itself was pioneered by Palantir over a decade ago. That is the whole market converging on one idea: AI needs customization, integration, and judgment to deliver value. This is AI's biggest opportunity and, increasingly, the market's baseline expectation. Off-the-shelf was always a myth, just like SaaS.
This summer, we've seen a spate of cutting-edge open-weight models from GLM, Kimi and DeepSeek. They're within a few points of the more expensive American frontier models, for a fraction of the price. And with CFOs demanding AI ROI, companies are becoming more comfortable leveraging these open-weight models, despite the added complexity of implementing model routing and the security concerns of using Chinese models. We expect this trend to grow in the coming quarters.
