Markets
Tech

Palo Alto Is Paying 45 Times Revenue for the AI It Says It Must Train Itself

Palo Alto Networks grew revenue 39% while GAAP operating income fell 69%, then agreed to buy Applied Compute for $4.5 billion in stock, roughly 45 times the target's annualized revenue. The same shares fund compensation, the CyberArk integration and now the AI lab.

Rows of server racks and bundled cabling inside a modern AI data center

Palo Alto Networks just posted the kind of quarter most security vendors would frame, and then priced an acquisition like a company expecting to pay in something scarcer than cash. Fiscal first quarter revenue came in at $3.44 billion, up 39% year-over-year, with Next-Generation Security ARR of $9.73 billion, up 66%, and remaining performance obligations of $21.0 billion, up 35%. The same day, it filed a merger agreement to acquire Applied Compute, a platform for post-training and inference of open-weight AI models, for $4.5 billion in Palo Alto stock, subject to adjustments. Applied Compute has more than $100 million in annualized revenue, reached within 18 months of founding, so the price works out to 45 times annualized revenue, paid in the buyer's own shares, with management guiding it to shave as much as 200 basis points off operating margin by fiscal 2028.

The quarter's fine print

The growth is real, and so is what it cost on the GAAP line. GAAP operating income fell to $96 million from $309 million a year earlier, a 69% decline that works out to a 2.8% operating margin on the quarter's revenue. Two items did most of the damage: share-based compensation of $622 million, up from $387 million, and amortization of acquired intangible assets of $282 million, up from $39 million, with $55 million of acquisition-related costs on top. Non-GAAP operating income was $1,059 million, up from $746 million, a 30.7% margin that rose 50 basis points. Gross margin of 75.2% fell 170 basis points. Net income of $482 million, or $0.57 per diluted share, was flattered by a $457 million income tax benefit that the release ties partly to a one-time adjustment on a historical transaction during the transition period of the Tax Cuts and Jobs Act. The diluted share count climbed to 838 million from 709 million. The company repurchased no shares in the quarter and carries a $1 billion buyback authorization through December 2027. Adjusted free cash flow was $1.9 billion, a 56.2% margin, down from 69.2% a year ago.

Buying the model it wants to police

The strategic claim is blunt. In the merger release, Palo Alto argues that no large provider is delivering those models securely and that it intends to be the partner that lets enterprises train, adapt and deploy open-weight models in public cloud or sovereign environments. On the call, CEO Nikesh Arora said open-weight models grew from under 5% of token traffic to more than 60% over the past year, and put the logic plainly: "I cannot go in and intercept and understand somebody else's application because I don't know how it's supposed to behave because I didn't train it." He said the company looked at nine different providers before choosing Applied Compute and expects to reach 100 frontier AI researchers by year-end. Applied Compute co-founder and CEO Yash Patil becomes Palo Alto's Co-Chief AI Officer after the close, expected in the fiscal third quarter of 2027 and subject to the Hart-Scott-Rodino waiting period.

What the price buys, and what it costs

Chief Financial Officer Dipak Golechha said on the call that the consideration includes about $500 million in cash and prepaid compute commitments, and that the deal should be accretive to top line growth from day one, add more than $1 billion of Next-Generation Security ARR by fiscal 2029, and lower operating margin by up to 100 basis points in fiscal 2027 and up to 200 basis points in fiscal 2028 before returning to the 29% to 30% range in fiscal 2029. Cumulative capital spending through fiscal 2030 should stay under $3 billion to build more than 100 megawatts of inference capacity. At $4.5 billion against $100 million of annualized revenue, the multiple is 45 times, or about 31% of the $14.545 billion midpoint of fiscal 2027 revenue guidance. At the $337.16 per-share reference price set by the 15-day volume-weighted average in the merger agreement, the consideration divides to roughly 13.3 million shares, about 1.6% of the 844 million to 847 million shares guided for fiscal 2027. Palo Alto had $9.1 billion of cash and investments at quarter end and still structured the deal entirely in stock. Stand-alone guidance holds non-GAAP operating margin at 29.5% on revenue of $14.52 billion to $14.57 billion, up 26% to 27%, with adjusted free cash flow margin of 38% and diluted EPS of $4.26 to $4.30.

The tension sits in the currency. The same stock funds employee compensation, the CyberArk integration and now the AI lab, while the GAAP margin keeps compressing under equity compensation and purchase amortization; goodwill already stood at $22.559 billion before this deal. The bet pays only if enterprises decide the vendor that trains and supervises their models should be the same one guarding their networks, and it pays on Palo Alto's schedule: the shares go out at closing, the margin dilution starts in fiscal 2027, and the promised ARR lands by fiscal 2029. The invoice arrives before the revenue does.