Akamai's Agentic Bet Runs Through the API Layer — The MuleSoft Pact Signals the Fight, Not the Finish

An AI agent rarely succeeds by breaking a model. It acts — it calls an API to pull up a customer record, trigger a payment, or kick off a workflow — and every one of those calls is a handshake between systems that has to be trusted. That is why, as the industry shifts from training models to deploying live agents, the security battleground is moving to the API layer. It is the same cycle I keep following, re-drawn: the contested ground stops being who trains the biggest model and becomes who can be trusted to govern each call an agent makes.

Akamai is making a focused bet that it can own that layer, and in June it announced the distribution vehicle. Akamai and Salesforce's MuleSoft said they are wiring Akamai's API security into MuleSoft's platform — API discovery, risk scoring, behavioral analysis, and security findings flow into MuleSoft's Agent Fabric, while MuleSoft-managed API context flows back to fill out Akamai's inventory. For joint customers, the integration becomes available in July.

The strategic read is straightforward. MuleSoft is the API-management gateway inside Salesforce's enormous enterprise install base, and Agent Fabric is where its customers register, route, and monitor AI agents. By plugging in there, Akamai gets its security product visible inside the exact workflow where agents get governed — and security is how a company that started as a content-delivery network hopes to become the connective tissue of the agentic era.

What the numbers actually delivered

The test of any positioning claim is whether it reaches a line item. On August 6, Akamai reported the quarter that ended June 30 — before the MuleSoft integration was even available. Total revenue was $1.10 billion, up 5% year over year. The security segment, Akamai's real engine at roughly 55% of sales, posted $604 million, up 10%. That sits alongside a Cloud Infrastructure Services unit growing 39% — to just $99 million — and a legacy delivery/CDN business down 6% at $396 million.

So the headline is better than it looks: security is the biggest line and it is growing. But 10% is not an accelerating-agent story yet; it is steady security growth in a market where AI-adjacent rivals are compounding far faster.

The price already noticed

Here is the tension a holder has to confront. The stock is up about 27% year to date and roughly 44% over the past twelve months, even while sitting near the middle of a $70-to-$165 range over that same year. It trades around 39 times trailing earnings. The agentic-API-security thesis is not undiscovered — the multiple is pricing it.

Margins tell the other half of the story. Akamai is spending heavily to build out the AI cloud infrastructure that is growing 39%: GAAP operating margin fell to 7% from 15% a year earlier, and non-GAAP margin fell to 25% from 30%. The growth engine is real, but it is expensive to build, and the company is guiding 2026 revenue to roughly $4.45 to $4.53 billion.

Pact versus proof

The honest reading separates the pact from the profit. The partnership is genuine and strategically sensible — it plants Akamai's security inside the governance workflows of Salesforce's enterprise customers, the kind of distribution that software-layer economics reward. But the agentic-specific revenue is a claim, not a delivered result; security grew 10% in a quarter that predates the deal entirely.

This is where I land. The long-term thesis is intact — agents will multiply API calls, and someone has to secure and govern them. The open question is whether the current return curve has already collected the good news, because at roughly 39 times earnings with shrinking margins, the market is not paying for 10% security growth. It is paying for acceleration that the numbers have not yet shown.

What would change my reading is one observable thing: security growth re-accelerating as agent attach spreads, ideally alongside stabilizing operating margins once the cloud buildout matures. Until those show up in a print, treat the MuleSoft pact as shrewd positioning with a real distribution advantage — and price the proof when it actually lands.