Akamai China exit. What AEM engineers need to know

By June 2026, Akamai will shut down its CDN operations in mainland China. For developers managing Adobe Experience Manager (AEM) platforms in the region, that’s not just a vendor switch. It’s a full-scale infrastructure reckoning.

Akamai has long played a foundational role in internet infrastructure, powering everything from e-commerce sites to global media platforms. Companies like Microsoft, Apple, Netflix, and Spotify have relied on Akamai to enhance content delivery efficiency and user experience It’s also earned recognition for its strong security capabilities, handling one of the largest DDoS attacks in history by defending against over 1.3 terabits per second (Tbps) of data traffic.

In China, Akamai edge servers helped international firms navigate the country’s unique internet ecosystem. But with the provider’s changing priorities, that edge is now disappearing.

The main reason, as is often the case with publicly traded companies, lies in the spreadsheets. Over the past year, Akamai financials have shifted dramatically. In Q4 2024, security and compute revenues surged to 69% of total revenue (comprising $535M in security and $167M in compute, while delivery revenue dropped 18% YoY). That reflects the company's stated transformation into "the cloud company that powers and protects life online”.

Akamai’s CEO, Tom Leighton, emphasised this pivot, noting that security and compute now make up two-thirds of its revenue and offer better growth prospects, leaving traditional CDN operations, particularly in China, as "dull by comparison".

But Akamai's departure is actually more multi-faceted. Beyond its strategic shift to cloud and security offerings, the company has cited the high regulatory burden of maintaining in-country infrastructure and operations in China.

The increasingly complex and demanding compliance requirements, coupled with macroeconomic and geopolitical headwinds, have made the China CDN segment less attractive for Akamai compared to its growing compute and security businesses.

Enterprises still relying on Akamai for their China-facing traffic, particularly those using AEM, need to act fast.

The exit plan was communicated in January 2025. With new provisioning halting from September 30, 2025, China’s CDN services will be fully decommissioned by June 30, 2026. 

After that, if nothing is done, traffic meant for Akamai’s China CDN will automatically be rerouted to nearby regions such as Hong Kong, Singapore or Tokyo. But latency doesn’t travel well. Estimates suggest a 150-300ms delay or more, potentially pushing Time to First Byte by over a full second for users in cities like Beijing or Shanghai.

And these are just metrics. The real-world symptoms will look more like pages timing out, videos buffering infinitely, and interactive experiences breaking down.

Akamai’s fallback plan involves transitioning China traffic via Tencent Cloud (which already supports Chinese domain names) and soon Wangsu Science & Technology, where Akamai will act as a reseller. But it’s just that - a handoff, not a turnkey migration.

Twelve months may sound like a generous buffer. For engineering teams, it’s not.

Replacing a CDN at this scale means more than flipping a switch. It’s a full-stack undertaking: new WAF rules, DNS migration, SSL handling, logic rewrites, and often complex contract renegotiations with local vendors. Like many modern CDNs, Akamai also underpins authentication, caching strategies, and content personalization at the edge. Each of those features needs to be rebuilt, tested, and validated, without breaking production environments. The complexity is compounded by vendor validation and contract negotiations, which can alone take months.

It’s no surprise that Akamai’s decision sparked heated discussion on Hackernews - one of the most active developer communities online.

In one forum thread, a developer working for an enterprise customer relying on Akamai noted that migrating off the platform would be far from simple. According to their comment, the transition would involve rebuilding pipelines, reconfiguring EdgeWorkers and edge caching rules, rewriting origin routing configurations, and likely addressing other dependencies that weren’t yet fully understood.

They estimated that fully executing the migration – while ensuring nothing breaks across the enterprise – would probably take about a year.

Another developer echoed this sentiment, explaining that in large enterprise software stacks, even seemingly simple tasks can take years due to hidden dependencies. They pointed out that China’s ICP filing process alone can take four to six weeks – further compressing the already tight timeline for compliance and deployment.

Waiting until 2026 to worry about these risks is a significant service outage. Akamai itself warns that customers "must complete the transition to our Partners’ solution by June 30, 2026, to maintain uninterrupted service".

Adobe does not provide a built-in solution for compliant content delivery inside Mainland China via AEM as a Cloud Service (AEMaaCS).

As noted in a thread on Adobe’s Experience League, AEMaaCS is not officially supported in China, due to both technical limitations (e.g., latency caused by the Great Firewall) and legal constraints. Adobe recommends that customers seeking to serve content within China explore using a China-specific CDN, such as Alibaba Cloud or Tencent Cloud, which can act as a BYOCDN layer pointing to the globally managed AEM Fastly CDN.

However, even this model falls short of full compliance unless content is served from within Mainland China, which AEMaaCS does not natively support.

In practice, some teams have adopted custom architectures – for example, Daniel Strmečki, a well-known contributor in the AEM community, chimed in on the same thread, describing a more custom alternative: hybrid architectures with local caching. "On my project, we are not using CDN for the China sites at all, as it is not required. Instead, we have implemented a Varnish cache in AWS China that grabs the content from the AEMaaCS instances hosted in Europe. So the final site is cached and delivered from China to align with the legal restrictions" he wrote, adding that "China editors use the same AEM author as all other editors worldwide, but the delivery of content to China users goes through AWS China instead of our standard Akamai CDN". 

This solution was marked as a "Correct answer". But let’s take a moment to explore some other possibilities.

For some, the short-term fix is a local CDN handoff partnering with a Chinese provider to host static content inside the firewall. But even this has limits. Traditional CDN models rely on pulling data from an origin. When that origin is blocked, slow, or foreign, performance breaks down.

That’s why some engineers explore more proactive architectures: push-based delivery models that pre-generate and distribute content updates to servers already located inside China. Instead of relying on runtime fetches from distant backends, this model pushes ready-to-render content to China-based caches or web nodes as soon as it’s updated.

This approach reduces latency, sidesteps firewall risks, and offers more resilience across dynamic and static content alike. For AEM users, it also opens the door to hybrid architectures: keeping AEM in Europe or the U.S., but syncing content to a China-optimized delivery layer. 

StreamX is currently recruiting companies and architects for a fast-track proof of concept, so you can test and validate that approach in a short project. Learn more here.

Akamai's departure from China is not an isolated incident. Other CDN providers like Edgio, StackPath, and Lumen have also scaled back or exited certain markets. There’s definitely a broader trend of digital fragmentation and increasing emphasis on digital sovereignty. Traditional tech models are reaching their limits in tightly regulated markets, and the era of "plug-and-play" Western cloud services in China is over.

AEM engineers should treat June 30, 2026, as a hard deadline to test, validate, and migrate architectures built for regulatory fragility, network segmentation, and performance resilience.

The clock is ticking.