Palo Alto Networks signed an agreement to buy Applied Compute on November 18, 2026, in the United States. The announced consideration is $4.5 billion in Palo Alto Networks common stock, subject to adjustments. The buyer expects closing in the third quarter of its 2027 fiscal year, subject to the applicable regulatory process and other closing conditions. Applied Compute runs infrastructure for adapting and serving AI models. Palo Alto is proposing to buy the company, its team and a business that charges for work closer to the creation and operation of models than its core security business typically does. Palo Alto Networks announcement · Merger disclosure on Form 8-K
The price is the first part of the economics. The rest appears on the earnings call, where management discussed capital spending, pressure on margins and a target for recurring revenue. The plan depends on an enterprise that already trusts Palo Alto to secure an application choosing it to help build and run the model inside that application. That may be a valuable connection between a sales channel and a product. It also gives the buyer a more active responsibility for the expense of running the product. An announcement can establish the price of getting in. The customers and the cost of serving them will decide how much the new business is worth to the owner. Palo Alto's fiscal first-quarter earnings call
What the agreement buys#
Palo Alto's filing says the proposed merger covers Applied Compute's outstanding stock, warrants and other rights to receive or acquire equity. The consideration is to be paid in common shares of Palo Alto Networks. The share calculation uses $337.16, the volume-weighted average closing price over 15 consecutive trading days ending November 17. The reference price is a term of the deal. The disclosure expressly says the number of shares to be issued at closing was still unknown. Securities issued to the target's eligible equity holders are expected to be issued to accredited investors under registration exemptions, according to the filing. Form 8-K, Item 3.02
Dividing the headline amount by the stated reference price illustrates the scale. The simple calculation comes to about 13.35 million shares before any adjustments. Palo Alto reported about 819 million common shares issued and outstanding at October 31, so that illustration is about 1.6 percent of the earlier count. It is useful to keep the reference date attached to both numbers. The actual issuance can differ, the buyer's own share count can change, and the filing describes the total consideration as adjustable. The arithmetic is a way to understand the size of a stock transaction without mistaking a division for the final share register. Merger terms · October 31 share count
The structure lets Palo Alto deliver part of the combined company's future value as payment while retaining cash for other purposes. The sellers, to the extent they receive and hold Palo Alto shares, have exposure to a public stock whose value can move before and after the deal closes. Existing shareholders can have a smaller proportional interest when new shares are issued. They may benefit if the asset expands the earnings and cash flow available to the enlarged group. How those effects balance depends on the final issuance and what the acquired business earns. Paying in stock is a form of capital allocation, with a different set of costs from funding an acquisition from the cash balance.
The 8-K also identifies the applicable waiting period under the Hart-Scott-Rodino Act as a closing condition. The company repeated the fiscal third-quarter timetable in its Form 10-Q. A signed agreement and a public disclosure are evidence that the parties intend to complete the transaction on the stated terms. The timing and the entitlement to acquire the asset still depend on closing. Yash Patil, co-founder and chief executive of Applied Compute, is expected to become co-chief AI officer at Palo Alto once the acquisition closes. The appointment belongs on the same conditional timeline. Closing conditions and leadership announcement · Subsequent event in Form 10-Q
From a venture round to a proposed sale#
Applied Compute announced on November 2 that it had raised $350 million at a $3.25 billion valuation in a round led by Elad Gil. The company said it had reached a $100 million annualised revenue run rate in less than 18 months after founding. It named established companies and younger technology businesses using its products. Those statements give a sense of the target that Palo Alto reviewed. A run rate takes activity over a shorter period and expresses it on an annual basis. The financing announcement does not establish that the company had already booked $100 million of revenue over the previous year, or what percentage of that figure it could retain as profit. Applied Compute's financing announcement
Two weeks later, the $4.5 billion announced merger figure stood $1.25 billion above the target's published financing valuation. The difference is about 38 percent of $3.25 billion. That comparison is useful as a quick way to locate the change in scale. A private funding round prices a particular investment with its own rights. A proposed acquisition sets terms for control of the company and covers existing equity and related rights. The disclosed figures leave little basis for working out the proceeds of each startup investor or employee. To do that would require their holdings, what they paid, the rights attached to the securities and the contract's allocation of consideration. Financing announcement · Acquisition disclosure
The two rounds of disclosure answer where capital goes in different ways. The November 2 release describes new funding raised by the target and says it planned to add compute capacity, hire and accelerate deployments. The merger disclosure explains what Palo Alto intends to transfer in exchange for the target's equity. Share consideration to selling holders is a way for those holders to exchange their ownership. It is separate from the future resources the owner chooses to put into the business. The fact that an institution appears in a funding round also says little about how much of the merger consideration it would receive. Keeping the two transactions distinct makes it easier to understand both of them.
On the call, Palo Alto said the acquisition includes approximately $500 million in cash and prepaid compute commitments owned by Applied Compute. Those items sit together in the amount management cited, without a breakdown in that passage. Cash can be used across operations. Prepaid compute is economically useful to the extent that the contracts supply capacity the business can use on suitable terms at the right time. The number is evidence that the buyer considered the assets and commitments it would inherit. Treating the whole amount as available cash or subtracting it mechanically to arrive at a final enterprise value would give a more precise answer than the disclosure supports. Palo Alto earnings call, remarks on the terms
The business inside the acquisition#
Applied Compute introduced its AC2 platform in August as a training and inference cloud for models with available weights. Post-training is the work of adapting an existing model to a task, policy or body of examples. Inference is the work of running the adapted model when requests arrive. A platform covering both stages can sell the resources for changing the model and the ongoing capacity to serve requests. Applied Compute's product material describes connecting future training to what a team learns from production. Commercially, that means the relationship can extend beyond the first experiment, if the customer chooses to keep running and improving the model there. Applied Compute's AC2 announcement
One disclosed project shows what that might look like. Applied Compute and Clay described building a router that selects which model should handle different types of requests in Claygent, a sales and marketing research tool. In a live A/B test cited by the provider, inference cost fell 47 percent relative to using a particular larger model for all those requests, while the named quality guardrails were maintained. The percentage belongs to that workflow, traffic and baseline. The case is useful because a purchaser can see what was being paid for. The benefit comes from dispatching an expensive task to a large model where it helps and using a cheaper one where it is sufficient. Applied Compute's Clay case study
The vendor's November financing release reported more than 1,700 customer training runs on AC2 since July and said 40 percent of its customers used both training and inference. Those are company-provided measures. They help explain the logic for keeping the two services together. The count of training runs says how often customers tried to adapt a model, without specifying the amount billed or how many runs were for sustained production. The overlap measure says that some customers chose both products. A buyer of the company still needs to learn the dollar mix, how usage changes as customers grow and what it costs to support the businesses operating those workloads. Applied Compute's operating disclosures
The economics of serving models make this different from adding another line to an existing software contract. Each inference request consumes computing time. Running an endpoint also takes memory, capacity and engineering whether demand is smooth or arrives in bursts. A vendor can improve the ratio of cost to useful work with better utilisation, routing and a choice of models. Its customer can also change which models it uses or reduce traffic. A contract with recurring spend can therefore carry a cost profile different from a standard software subscription. The revenue a combined company can build matters. The portion it can keep after serving the demand matters too.
Why a security vendor wants the job#
Palo Alto says enterprises will use a mix of models whose weights are available and closed models offered through an API or service. It plans to bring secure training and inference to public and sovereign cloud settings, while using the research team across its own product line. The company's security tools typically help organisations govern data, software and access within a particular service or infrastructure. This acquisition is an attempt to get involved earlier, when a customer is deciding how its model will be adapted, hosted and maintained. If the two products work together, the company can offer both capacity and a reason for the security team to trust how that capacity is governed. Palo Alto acquisition announcement
That ambition has two distinct outlets. Palo Alto can use Applied Compute's people and tools to improve the models inside its own products. It can also sell the platform and services to organisations building models for their own use. The first may add features or change the cost of delivering an existing service. The second needs customers to put work and budget into a new category from their security vendor. The release describes both, and the company says Applied Compute will continue to work with existing kinds of customers. Internal adoption can provide evidence of the technology working in one environment. Commercial adoption asks whether other organisations choose it under their own constraints. Acquisition release and statements by management
A sovereign deployment is an example of the difference between a sales opportunity and a delivered product. A customer can ask where training inputs, checkpoint files and operational records are kept, who operates the infrastructure and which service obligations apply in that region. Model weights that can be held and run in a chosen environment can make it easier to design around some constraints. The customer still has to check the terms attached to those weights and the service, and what happens to training data and model versions. Palo Alto's announcement gives its intended scope. Individual contracts and deployed services will provide the evidence of how well it meets those needs.
There are two sides to the channel argument. Managers at an existing customer may welcome a security vendor that can take responsibility for more of a deployment. A buyer with an existing cloud or model supplier may prefer to keep the model layer open to its own choices while buying security separately. Those preferences matter once the new product is sold. Bundling might save a customer time on procurement. It might leave the customer with a more involved decision about portability and the cost of switching. What matters is the degree to which the platform makes each choice easier to carry out. Size of the sales force is a reason to test that proposition, and customers keep the right to decline it.
The target that has to be earned#
Management said on the earnings call that it expects Applied Compute to contribute about $1 billion or more to next-generation security annual recurring revenue by fiscal 2029. That is a forecast under the buyer's metric. The target's published $100 million figure is an annualised revenue run rate from a different source and period. The similarity of the units makes it tempting to put both numbers on a straight growth line. The definitions, products covered and count of renewing customers matter as much as the endpoints. One useful question for future filings is how much of the stated target reflects recurring contracts that are already running, additional usage charged to those customers and sales of new deployments. Palo Alto earnings call · Applied Compute financing release
Palo Alto gives readers a definition in its quarterly filing. The metric annualises the revenue allocated to eligible active contracts on the reporting date, excluding hardware and certain legacy or service offerings. The company reported about $9.7 billion on that basis at October 31. Contract dates, renewals and what products the measure includes can all change the total. An increase in the measure can help track a newer group of products. The timing of reported revenue and the value of future performance obligations remain different questions. This is especially relevant if Applied Compute's future contribution includes packages that also contain other Palo Alto products. The way management attributes the annualised value would affect comparisons over time. Palo Alto's definition of next-generation security ARR
The call put a second obligation next to that target. Management expects cumulative capital expenditure needs of less than $3 billion through fiscal 2030. Assuming a third-quarter fiscal 2027 closing, it forecast an impact of up to 100 basis points on the operating margin in fiscal 2027 and up to 200 basis points in fiscal 2028, followed by a return toward its stated margin range in 2029. These are management expectations and use its margin framework. They illustrate why the amount paid to acquire a business and the capital it needs to grow belong on different lines. The extra expense can come before there is enough demand to use all of the added capacity. Palo Alto earnings call, financial outlook for Applied Compute
The useful progression would be from a model that works in a pilot, to a service that customers can run with the required controls, to usage or commitments that recur at a manageable cost. Contract retention would show whether the combination helps enough for a customer to keep it. Usage and the expense supporting it would show whether the platform gains scale on the supplier's side. The company can give a better account of those steps by disclosing what sales include and which costs relate directly to serving customers. A year with fast revenue growth can still have uneven economics if the business has to reserve an infrastructure footprint well ahead of demand.
The announcement came with a financial report#
The acquisition was announced with Palo Alto's fiscal first-quarter results for the three months ended October 31. The company reported $3.445 billion of revenue, about 39 percent more than in the same quarter a year earlier. Its 10-Q attributes growth to greater adoption of its portfolio and contributions from recent acquisitions. The agreement to acquire Applied Compute was reported as a subsequent event. The quarter ended before the deal was signed. The day's results and the deal provide two kinds of information, about an existing business and about a new one management wants to add. Official results release · Form 10-Q
There is a similar boundary in the guidance. Palo Alto's chief financial officer said the issued guidance excludes the pending Applied Compute acquisition. It puts fiscal 2027 revenue between $14.52 billion and $14.57 billion on that basis. The $1 billion acquisition-related figure is a different metric on a different timetable, so adding it to the revenue range would even change the meaning of the numbers. For a later comparison, the useful questions will be when the reporting scope changes and whether the company can show the growth of the business it owned before the deal separately from contributions of the business it buys. Earnings call, guidance discussion
Reuters covered the proposed acquisition alongside the change in annual guidance, and the Associated Press mentioned both the transaction and the quarter in its market report. Their reporting places the deal in the context in which the public heard about it. The company was simultaneously giving information on the current business, its forecasts and the acquisition. An observed share-price move in that context belongs to the bundle of information reaching investors. Assigning it solely to a promise about the acquisition would require more evidence. The transaction arguments hold on their own merits even if the first reaction was to the results. Reuters report · Associated Press market report
The purchase follows earlier completed acquisitions that give a useful comparison for reading later disclosures. Palo Alto recorded $2.951 billion of purchase consideration for Chronosphere after it closed on January 29, 2026, in its fiscal 2026 annual report. That amount includes cash and part of the fair value of replacement equity awards. Other portions of awards were allocated to services to be performed later. The example shows how a closing and a purchase accounting note can provide details that a signing announcement leaves open. It says nothing about the allocation or the ultimate expenses for Applied Compute. Those details will become measurable only as the later filings report the facts of this deal. Palo Alto fiscal 2026 annual report, Acquisitions note
What will settle the argument#
The next document that can change the account of the transaction is a closing announcement with actual consideration and terms, or a filing explaining changes to the agreement. A later set of accounts can show when the target enters the results and how the buyer allocates the purchase price across the assets and liabilities it receives. Those facts help connect a stock payment to the economics of the group. There is also a product milestone with its own deadline. Palo Alto will need to show that customers besides the target's original accounts can buy the platform, understand the location and controls of a deployment, and see the costs across training and use.
The most persuasive evidence for management's plan would be real adoption in the existing customer base, accompanied by disclosure that helps a reader track the returns on the extra computing investment. Continued contracts with Applied Compute's original customers would also indicate whether the product has kept its strengths during the transition. A pattern of upfront capacity spending without evidence of repeat use would weaken the argument. So would a model service that customers can test in one setting but cannot bring into the environments where their sensitive data or established contracts require it to run. Both are commercial questions that can be tested over time.
Palo Alto has put a price on the equity it plans to acquire. Its filings also offer a useful starting point for deciding what counts as evidence afterwards. Shares delivered at closing will show what the owners exchanged. Spending and margin disclosures can show the cost of extending the business. A recurring revenue figure and the contracts beneath it can show whether companies are buying that extension. The strategic part of the deal is choosing to provide security and the model infrastructure together. The financial test will be whether customers keep paying for that combination in enough volume to justify offering it.