Check Point Declares the Trough Again — And Hikes Firewall Prices While Demand Is Falling
Q3 revenue fell 1%, growth-tech ARR rose 38%, and a 12–20% price increase is now the bridge to 2027. The market may not care about either.
CHKP · Earnings Call · 2026-11-02
The word 'trough' is doing a lot of work
Check Point's third quarter was, on the surface, a non-event: total revenue of $670 million, down 1% year-over-year at the midpoint of guidance, non-GAAP EPS of $2.49, adjusted free cash flow of $258 million. The growth engine inside the flat number is real — subscription revenue of $339 million, up 11%, now 51% of the total, with the trio of email security, exposure management and AI security growing ARR more than 38% and calculated billings more than 47%. Email security alone crossed $200 million in ARR.
But the headline the company wants you to take away is a claim about the future, not the quarter. CFO Roei Golan: “we believe Q3 revenues represent the trough, and we expect to see the beginning of the improvement starting in Q4.” — Roei Golan, CFO or Finance Executive · 2026-11-02 CEO Nadav Zafrir then immediately qualified it: “the pace of recovery is more gradual than we previously anticipated.” — Nadav Zafrir, CEO · 2026-11-02 Guidance is $720–770 million for Q4 — flat year-over-year at the midpoint — with the full year landing at $2.732–2.782 billion.
The trough framing is not new. It is the third consecutive quarter in which the bottom has been declared and the recovery has been pushed slightly to the right.
We are no longer a firewall company — except when we need one
Zafrir's rhetorical pivot is striking: “we're no longer a firewall company, and we're no longer a network security company.” — Nadav Zafrir, CEO · 2026-11-02 The strategy is preemptive security — an agentic management layer that simulates attackers, a new AI network firewall, and an agentic security management platform that translates business intent into Zero Trust policy across multi-vendor estates.
Yet the mechanical bridge to 2027 is the oldest business they own. On October 4 Check Point raised firewall prices 12–20%, including attached support and subscription. Golan quantified the payoff — “it can be up to three points equal to three points to our growth... historically, we managed to benefit slightly less than half of it” — Roei Golan, CFO or Finance Executive · 2026-11-02 — and tied the decision explicitly to memory cost inflation, which clipped gross margin to 87%. Analysts noticed the tension instantly: raising price into a double-digit product decline is an unusual move, and management's defense was that competitors did it at the same time. Fortinet, reporting in the same window, was talking about the same AI security demand and the same component lead times. That is a shared, sector-level pricing action, not a company-unique insight.
It really sounds like it's been hard for you guys to capture new customers. By the way, there's this sort of perception of Check Point as a really good technology company, but sort of a legacy technology company.
That was John DiFucci asking the question of the call. It is the same question, rephrased, that has been asked for over a year.
The prior calls say the word 'execution' a lot
In July 2025, Tal Liani put it bluntly: “we're not seeing acceleration.” — Tal Liani, Analyst · 2025-07-30 By February 2026, Zafrir's answer had compressed to “we laid the foundations. Now it's all about execution.” — Nadav Zafrir, Chief Executive Officer · 2026-02-12 By April 2026 the go-to-market reorganisation was described as “sort of a blip but it does take a little time” — Unknown Executive, Executive · 2026-04-30, with a promise of improvement in the second half. This quarter the explanation is more precise — the qualified funnel is still better than last year, “but the magnitude is slightly lower” — Roei Golan, CFO or Finance Executive · 2026-11-02, because large refresh projects with existing customers slipped into 2027.
That is an important distinction. This is not a demand problem dressed as a supply problem; it is a timing problem management has now forecast incorrectly three times in a row. The offsetting lever is headcount: dozens of new sellers onboarded, more signing, with the stated goal of entering 2027 with materially higher sales capacity. Sales capacity has been a recurring promise since at least the July 2025 call. It is now the load-bearing element of the whole 2027 story.
The genuinely new thread: sovereign AI data centres — into a tape that is selling them
Buried in the Q&A is the most interesting strategic shift. Asked about firewalls protecting AI data centres, Zafrir turned expansive: “the investment in data centers, and specifically AI data centers, whether it's for training or for inference, is just mind-boggling.” — Nadav Zafrir, CEO · 2026-11-02 He then connected it to data sovereignty, open weight models run locally, neoclouds competing with hyperscalers, encrypted east-west traffic that needs stateful inspection, and a deepening collaboration with NVIDIA at the hardware level. The new Maestro Hyperscale Orchestrator is aimed squarely at that use case.
This is a new keyword cluster for Check Point — sovereignty and open-weight deployment did not exist in these calls a year ago, and it is the one place where the company's 30-year on-premises heritage becomes an asset rather than a liability. It also, conveniently, reframes data center security as incremental revenue starting 2027.
The problem is the tape. The AI data center theme is one of the most heavily sold clusters in the market across both the 30- and 90-day windows, with the memory complex at the epicentre — high bandwidth memory down roughly 30% over 90 days, and Sandisk this week talking up inference memory architectures and KV-cache. Check Point is asking investors to pay for exposure to an AI infrastructure theme at precisely the moment that theme is being de-rated — and simultaneously citing the same memory inflation as a gross-margin headwind. It is a genuinely awkward juxtaposition: the demand driver and the cost headwind are the same phenomenon.
What actually protects the equity
The bull case here is not the growth narrative; it is the cash. Adjusted free cash flow of $1.179 billion over the trailing twelve months, roughly $4 billion of cash and equivalents, and a buyback running at 2.5 million shares for $325 million in the quarter at an average of $132 — 9 million shares and $1.4 billion over twelve months. An analyst on the call noted the stock trades near a 10% free cash flow yield; public coverage puts the shares around $133 and down about 32% over the past year, against a market capitalisation near $14.4 billion.
Management declined to lean harder, saying the buyback is reviewed quarterly with the board and unchanged for now. With roughly $50 million of incremental FX headwind already flagged for 2027 operating expenses, a 12–20% price rise that historically monetises at less than half its face value, and an Security for AI business that is still small enough that its standalone contribution is described as future-tense, the balance sheet is doing more of the work than the P&L.
The story, then, is not 'trough reached.' It is a company whose network security franchise is being repriced upward to defend a margin, while its growth narrative is being re-anchored to an AI infrastructure theme the market is currently punishing. Both can be right. Neither is yet visible in the numbers.