AWS bulletins fix a mount-option injection flaw in the Amazon EFS CSI Driver for Kubernetes (versions 3.1.0 to 3.4.2, fixed in 3.5.0) and arbitrary command execution when GluonTS loads untrusted model directories (before 0.17.0).
Palo Alto Networks to acquire Applied Compute in a US$4.5 billion all-stock deal
EditorialBy TrustList Editorial
Palo Alto Networks agreed on 18 November 2026 to acquire Applied Compute, a post-training and inference platform for open-weight AI models, in an all-stock deal valued at US$4.5 billion, to close in its fiscal third quarter of 2027.
About Palo Alto Networks to acquire Applied Compute in a US$4.5 billion all-stock deal
Palo Alto Networks to acquire Applied Compute in a US$4.5 billion all-stock deal
18 November 2026: Palo Alto Networks has signed a definitive agreement to buy Applied Compute, which runs a platform for post-training and serving open-weight AI models. The cybersecurity company will pay in shares, in a transaction valued at US$4.5 billion and subject to adjustments, and expects to close in its fiscal third quarter of 2027, once regulators have cleared it and other customary conditions are met.
Not yet independently verified. The deal value, structure, expected closing and quotes come from Palo Alto Networks' own release of 18 November 2026, read on 19 November. Applied Compute's own announcement was not read. The value is stated as subject to adjustments, and no exchange ratio or price per share was given. We will update this when it can be confirmed, and remove this note.
What Applied Compute does
Palo Alto Networks calls Applied Compute "a leading platform for post-training and inference of open-weight AI models". Post-training is the work of adapting a published base model to a company's own data and tasks; inference is running the adapted model in production. Applied Compute serves AI-native companies, digital natives and large enterprises, and will keep doing so after the deal, according to the release.
Applied Compute recently raised US$350 million at a US$3.25 billion valuation, so the deal values it well above that round.
Why Palo Alto Networks wants it
The plan is to let enterprises train, adapt and deploy open-weight models with security built in, in public clouds or in sovereign environments where data must stay inside a country or a company. Chairman and chief executive Nikesh Arora said: "Applied Compute has built the best post-training and inference platform we've seen." The release says Palo Alto Networks saw the platform first hand as a prospective customer, and that its Unit 42 Frontier AI Assessment confirmed the startup's security foundations.
Applied Compute's team will join Palo Alto Networks, and co-founder and chief executive Yash Patil becomes Co-Chief AI Officer after closing. Patil said joining "will put security at the center of all our products."
The company said it would give more detail on its fiscal first quarter 2027 earnings call, held on 18 November.
What it means for buyers
For security teams, the deal is another sign that the large security vendors want to own the AI stack, not only protect it. Customers of Palo Alto Networks can expect model hosting and adaptation to appear alongside its existing AI security products, with the pitch that one vendor covers both.
Current Applied Compute customers should ask how pricing, data handling and support will change between signing and closing, and afterwards, and whether the platform will keep supporting customers that do not use Palo Alto Networks products. Companies that chose open-weight models to avoid lock-in to one AI provider should check whether running them on a security vendor's platform brings a different kind of lock-in, and keep their fine-tuning data and adapted model weights exportable.
Until the deal closes, the two companies remain separate, and nothing changes for customers on either side.
Sources
Categories & features
TrustList Weekly
The week in software and IT, in one email
The news that matters to buyers, new rankings and our own research. Every Thursday, free, and easy to leave.
You may be interested in
AWS bulletin 2026-125 fixes a flaw in SageMaker Unified Studio Space start-up that could run code in another project member's Space. Supported images are patched on restart; 2.8–2.13 and 3.3–3.8 are end of support and get no fix.
Nvidia reported third-quarter revenue of US$115.8 billion, up 103% on the year, forecast US$130 billion with no China data centre compute revenue assumed, and created a board committee to oversee national security and export controls.
Boltz, the biomolecular AI company that began as a group of MIT researchers, announced a US$110 million Series A led by Bessemer Venture Partners on 18 November 2026, saying it works with six of the 20 largest drug companies.
Anthropic priced its initial public offering of 980 million Class A shares at US$83.00 each on 18 November 2026, with trading on the Nasdaq Global Select Market expected from 19 November under the symbol ANT and closing on 20 November.
San Francisco generative media platform each::labs says it raised a US$7 million seed round co-led by Revo Capital and the EBRD on 18 November 2026, after its annualised revenue run rate passed US$20 million in October.
The start-up lists Stanford professor Yejin Choi and co-chief executive Prithviraj Ammanabrolu on its team; it has not published the investors or the stage itself.
The San Francisco startup, from Y Combinator's Spring 2026 batch, builds patient-level datasets from scattered hospital records and works in the US, Brazil, India and Argentina.
More on TrustList
Everything here links back to the same verified catalogue. Pick your next stop.
- More United StatesThe ranking for this subject
- CompaniesAgencies, consultancies and IT service providers, ranked by verified reviews.
- ProductsSoftware and SaaS with pricing, features, integrations and alternatives.
- AwardsAnnual recognition decided by verified reviews and an independent jury.
- LaunchesNew products and releases, voted up by the community every day.
- AI ModelsBenchmark scores and community ratings for every major model.
- RequestsBuyers describe what they need; vendors respond directly.
- PeopleReviewers, authors and makers with public profiles.
- ComparePut up to four listings side by side before you shortlist.