Palo Alto Networks to buy Applied Compute for $4.5B in stock, WSJ reports
Palo Alto Networks has agreed to acquire Applied Compute, a startup that helps companies customize open-weight AI models, for $4.5 billion in stock, the Wall Street Journal reported. Applied Compute was valued at $3.25 billion earlier this month, so the purchase price represents a substantial premium over that private valuation. The company was founded last year by former OpenAI researchers and specializes in tuning open-weight models — models whose trained weights are published for download, so customers can adapt them to their own data rather than only calling a vendor’s hosted API. The report describes an agreement to buy; it gives no closing date or regulatory timetable.
Verified and unconfirmed
- Verified: The Wall Street Journal reported that Palo Alto Networks agreed to buy Applied Compute for $4.5 billion in stock, that the startup was valued at $3.25 billion this month, and that it was founded last year by former OpenAI researchers to customize open-weight AI models.
- Not yet confirmed: No closing date, financing condition, or regulatory review timeline is reported, and neither company’s own statement is cited in the account. The report also does not say how Applied Compute’s customization tools will be integrated into Palo Alto Networks’ security products, or whether existing customers will keep the same standalone service.
Background
Open-weight models are AI systems whose trained weights are published, so organizations can download, host and fine-tune them instead of relying on a closed vendor API — but adapting one to a specific business use still requires specialist engineering. Applied Compute, founded last year by former OpenAI researchers, sells that customization work to companies, which is what the cybersecurity vendor Palo Alto Networks is buying into; the Wall Street Journal account does not describe the customer base, revenue, or how the technology would be folded into Palo Alto’s products.
Why it matters
Companies that rely on Applied Compute to fine-tune open-weight models now face a vendor-ownership question: the startup’s customization work is being folded into a cybersecurity vendor whose platforms are sold to security teams, not model builders. Until the all-stock deal closes, those customers should confirm in writing how their fine-tuned weights, training data and tooling can be exported if the combined product roadmap shifts toward security use cases.
What to watch
Whether the all-stock deal clears regulatory review and closes, and whether Palo Alto Networks discloses the purchase terms in an SEC filing or earnings call that confirms the price independently of the Journal’s reporting.
Comments
Comments use a GitHub account and are moderated. Comment rules