Edge Development Platforms

In October 2021, Forrester published its first evaluation of edge development platforms and named two Leaders. In March 2026, it published its second. The same two vendors sit on top. Nothing else about the report is the same.

The Forrester Wave: Edge Development Platforms, Q1 2026 is out, with vendor announcements beginning March 24. The stability at the top is the story the rest of the scorecard keeps contradicting: a heavier format, a wider scorecard, a reshaped field, and criteria that did not exist when the first edition shipped.

Nine vendors were evaluated. The 2021 debut was a New Wave, Forrester's format for emerging markets. The 2026 edition is a full Wave. A promotion is a verdict: the analyst now treats edge development as an established market, and the same two names, Cloudflare and Fastly, are again the only Leaders. They are not tied.

The second edition, and the podium that refused to move: The Forrester Wave: Edge Development Platforms, Q1 2026

The report's framing sentence is worth repeating exactly. Edge development platforms, it says, "are becoming the primary interface for building responsive, sovereign, and cost-efficient software." The operative shift is workload affinity: placing compute and logic next to the users, the data, and the AI interactions, with the edge as the happy path for latency-sensitive workloads.

The placements that are public tell a clear story.

Cloudflare scored the highest possible marks in twelve criteria, including developer experience, initialization and deployment, state and storage, scaling, performance and latency optimization, security and permissions, and AI application development, and it took the top score in the strategy category outright. Vision, innovation, and roadmap were all maximums. The cited mechanics are concrete: zero cold starts, automatic workload placement, a 100 percent uptime SLA, an AI Gateway for model routing and governance, and an emerging agent-first developer ecosystem built on an Agent SDK and Sandboxes. The Replicate acquisition deepens the model catalog and GPU coverage extends it. Forrester's fit line: a strong fit for organizations that seek edge-first applications, APIs, and performance-critical user experiences, with momentum that increasingly extends into AI-driven workloads.

Fastly received five out of five in eight criteria: innovation, developer experience, initialization and deployment, performance, latency optimization, scaling, observability, and workload and network isolation. Its strategy centers on a WebAssembly-based global compute fabric built for deterministic performance, with microsecond startup times and request-level isolation. One more detail separates it from the rest of the field: Fastly is the only evaluated vendor to receive above-average customer feedback. Customers highlight strong performance and reliability, and the report credits its highly engaged technical support and proactive partnership. The fit line: a top choice for performance-critical, event-driven, security-sensitive, and globally consistent edge workloads.

Akamai is a Strong Performer, the strongest case for scale in the field: more than 4,300 edge locations and 41 core cloud regions, deep security integration across WAF, bot management, and Firewall for AI, the recent Fermyon acquisition bringing sub-millisecond WebAssembly execution, and workload flexibility through EdgeWorkers, EdgeKV, and distributed containers. Forrester positions it for planet-scale, performance- and security-sensitive workloads.

Amazon Web Services is a Contender, evaluated on CloudFront Functions and Lambda@Edge. The report calls it a natural fit for enterprises already deeply invested in AWS, then names the cost: it lacks the unified developer experience of specialized edge platforms. The hyperscaler treats the edge as an extension of the legacy cloud. The edge-native vendors treat it as the point.

The rest of the field is named in vendor materials: Netlify, Vercel, Azion, Microsoft, and Blaize. Their individual tiers sit behind the paywall, and the report's author was not publicly confirmed at the time of writing.

What the 2021 New Wave recorded

The first edition was The Forrester New Wave: Edge Development Platforms, Q4 2021: nine vendors, ten criteria, and a market still being defined.

Cloudflare led it on Workers, with the highest possible scores in developer experience, programming model, platform execution model, integrations, and day two-plus experience. The fit line then: the best fit for companies that need global support for web and content workloads.

Fastly led on Compute@Edge, differentiated across six criteria including vision, roadmap, developer experience, platform execution model, and day two-plus experience, and the only vendor in the field with a differentiated security score. The detail Forrester chose to quote is the one everyone remembers: the runtime starts within the time it takes to complete a TLS handshake, so pre-warming for serverless performance becomes unnecessary.

StackPath was a Strong Performer, the best fit for companies that need more options at the edge, running containers, virtual machines, and serverless scripting. It exited the market in 2023. Akamai was in the 2021 field without a Leader spot. Azion was there too.

The scorecard grew around those ten criteria. In 2026 a single Leader maxed twelve criteria on its own. Cold starts were the argument in 2021. In 2026 they are an entry requirement, and the argument is AI.

One scorecard, nine different products

The nine names in the 2026 field do not form a market. They form a Venn diagram.

Three are CDN natives turned compute platforms: Akamai, Cloudflare, Fastly. Two are hyperscalers reaching outward: AWS and Microsoft. Two are web development platforms: Netlify and Vercel. Azion is an edge platform in the direct lineage. And Blaize is an AI silicon vendor. A chip vendor in a developer platform Wave is the clearest signal of where the category thinks it is going: edge development is being defined by where inference will run, not just where web requests land.

That composition explains the Contender verdicts better than any feature comparison. The hyperscaler's edge is an extension of an existing estate, scored against platforms whose entire product is the edge. The web platform's edge is a deployment target, scored against platforms whose product is placement. Nobody in the field sells the same thing as the vendor next to them, and one ruler scored all nine.

The two Leaders are not tied

Forrester gives no shared first place. Two Leaders, opposite profiles.

Cloudflare wins the scorecard. Twelve maximum criteria, the top strategy score, the most complete AI story in the field: model routing through AI Gateway, an agent SDK, sandboxes, a model catalog, GPUs. A buyer betting on where the category goes next finds the roadmap already scored at maximum.

Fastly wins the references. It is the only vendor with above-average customer feedback, and the report quotes customers praising not just performance but the quality of the technical relationship. Its architecture is the narrower bet: WebAssembly isolation, deterministic performance, microsecond starts. A buyer betting on production behavior today finds the proof in someone else's invoices rather than in a roadmap slide.

Two Leaders. One wins the roadmap. The other wins the references. The tier names them equal. The criterion scores do not.

The word the new scorecard added: sovereign

The 2026 framing sentence contains a word that had no business in a 2021 edge computing report: sovereign. Edge development platforms are now positioned as the primary interface for building responsive, sovereign, and cost-efficient software.

The word does not appear by accident. When compute and data sit at the edge, the edge is a jurisdiction, not a cache tier. Workload affinity means user data lands in a specific country. That is a data residency decision wearing a latency argument, and the report has started scoring the platform on it.

For European buyers this is the part with legal teeth. Sovereignty is easy to claim in marketing and hard to deliver in architecture: data residency, key handling, the jurisdictions your traffic can pass through, what a US-based vendor can produce under a lawful access order. The scorecard has the word. The data processing agreement still has to have the mechanics. Buyers should get sovereignty in writing, at the same table where they get the uptime SLA, which by contrast is now advertised as a given.

What the criteria leave unmeasured

Here is the honest limitation of the scorecard, and it follows from the Venn diagram.

The criterion list is written in CDN language. Zero cold starts, initialization and deployment, performance and latency optimization, global workload placement. All of it assumes the edge is a point of presence near a web user. A buyer whose edge is a factory floor, a branch office, or an offshore platform will find the ruler measuring the wrong object. The report scores global reach beautifully. It cannot tell you whether the platform reaches your floor.

Tier stability has a shadow. The podium did not move, which is a settled answer about the edge-native architecture. But tiers do not price switching costs. AWS scores as a Contender and may still be the rational purchase for an enterprise whose estate, staffing, and compliance story are already AWS. The Wave ranks products. It does not rank the cost of leaving the thing you already have.

And the AI criterion is a photograph, not a promise. AI application development was maxed by the strategy Leader in March 2026, at a moment when edge AI is mostly inference placement and agent orchestration over APIs. The criterion is real. The practice is young. A maximum score today tells you the roadmap is built. It does not tell you what breaks when your inference volume grows tenfold.

Three questions before you follow the podium

Where is your edge actually located? If it is a point of presence serving web users, this scorecard was written for you. If it is a factory floor, a branch office, or a vessel, ask the vendor to show you a production deployment that looks like yours before you believe the scores.

Which gap can you live with: roadmap or references? One Leader maxes the strategy scorecard. The other is the only vendor whose customers report above-average satisfaction. Decide which of those your organization will actually be held to in year two.

What do AI and sovereignty mean in writing? Model routing through a gateway is not the same as inference at the point of use. And sovereignty needs mechanics, not adjectives: residency, jurisdiction, key handling, lawful access. Get both sets of answers attached to the contract, not the slide deck.

The podium did not move, and that is the most useful fact in the report. A market that keeps the same two Leaders across five years and a format promotion has answered its central question: the edge-native fabric won. What remains open is everything around that answer. The scorecard scores platforms. It cannot score your workload, your geography, or your auditors.

Engineering throughput on this platform still shows up in the same measurement layer covered in Software Engineering Insights Solutions, where portfolio-level value flow and individual-developer tracking are, per Forrester's own criteria, two different products wearing one name.

Forrester does not endorse any vendor named here, and tier placement should not be read as a recommendation to buy.

This market sits next to Edge Distribution Platforms, covered separately on this site. Gartner's most recent guide for this category is over two years old, and it already found that only 19 percent of enterprises had deployed edge computing, the least mature of the market's three core capabilities.