Vercel’s $150M Series D in November 2021 stated the thesis out loud: build “the end-to-end platform for the modern Web.” The acquisitions that turned that thesis into product were financed later, by a $250M Series E in May 2024 and a $300M Series F in September 2025. The platform a team adopted in 2021 as a thin Next.js host now wants to sell them analytics, backends, and an agent runtime, and the question for that team is whether the convenience is worth the exit cost.
What did the Series D actually promise?
The Series D promised a single-vendor platform. The announcement framed the round as funding “the end-to-end platform for the modern Web,” prioritizing infrastructure “brought into a single workflow” and “the best developer experience for every part of the software lifecycle,” according to Vercel’s own post. That is not a “faster deploys” pitch. It is a whole-lifecycle pitch, written in 2021.
The round added $150M at a valuation above $2.5B and brought Vercel’s total funding to $313M, with GGV Capital leading and a long syndicate including Accel, Greenoaks, GV, and Tiger Global participating, per the company’s announcement. Valuation tracks how well the thesis landed with investors. The Series D priced Vercel above $2.5B. The Series E in May 2024 closed at $3.25B, and the Series F in September 2025, co-led by Accel and GIC, landed at $9.3B, per the company’s funding history. By the Series F, the bet was $9.3B deep, and Dealroom lists Vercel’s total funding at $863M.
Which stack layers has Vercel pulled inside the platform?
Vercel’s acquisition trail maps onto the layers a frontend team used to buy independently: build tooling, analytics, and now agents.
| Tool | Layer absorbed | Date |
|---|---|---|
| Turborepo | Build / monorepo tooling | Dec 2021 |
| Splitbee | Analytics | Oct 2022 |
Turborepo appears in Vercel’s GitHub org, and Dealroom’s company record logs a December 2021 Vercel acquisition. Vercel acquired Splitbee on October 25, 2022, folding its page-view and visitor tracking into Vercel Analytics.
The Splitbee deal shows the mechanism that matters. When Vercel folded Splitbee into Vercel Analytics, the integration was described as “architectural, not cosmetic”: because Vercel controls both the hosting runtime and the edge network, collection happens at the infrastructure layer and was positioned as a “core product differentiator, not a bolt-on,” per Groundy’s earlier analysis. That distinction is the whole argument. A bolt-on you can replace. Infrastructure you cannot migrate away from cleanly.
What is “Agentic Infrastructure,” and why is it on the homepage now?
Vercel’s homepage now leads with “Agentic Infrastructure,” pitching agents that “reason, execute code in isolation, run for hours, and recover from failure,” with named features Durable Orchestration, Sandboxed Environments, an AI Model Gateway, and Fluid Compute, per the current homepage.
The product surface has reorganized into three tiers: Agents, Apps (marketing sites, SaaS backends, storefronts; Zapier is cited at 100M+ monthly visits), and Platforms (multi-tenant products with tenant isolation and custom domains; Mintlify at 20,000+ companies). That is a long way from “deploy your Next.js app,” and it is the shape the Series D thesis always pointed at.
The runtime work that enables the rebrand predates it. Fluid Compute is one of the named agent features on the current homepage, and it is the layer the agent pitch depends on: agents that “run for hours” do not fit a request-response serverless billing model. The rebrand is branding; the runtime layer is what makes it plausible.
Vercel’s GitHub org confirms the direction. It now hosts “workflow” (a Workflow SDK for durable, observable agents in TypeScript), “eve” (“The Framework for Building Agents”), and “ai” (the AI SDK), per the org page. The orchestration and framework layer is moving inside Vercel rather than staying a third-party concern. Notion is cited as powering millions of agent conversations daily on Vercel, which is the customer logo that turns the pitch from slide-deck claim to shipped workload.
Where does the lock-in actually sit now?
The lock-in surface has moved past static delivery. A team that adopted Vercel to run next build now has its CI previews, observability data, runtime, and agent orchestration on the same billing line, and each layer is data that does not follow it out the door.
The analytics critique generalizes to every absorbed layer. Because Vercel Analytics collects at the infrastructure layer rather than through an injectable client script, the data lives in Vercel, and a team that moves hosting providers leaves its historical analytics behind, per Groundy’s Splitbee analysis. Swap analytics for preview-deployment history or agent run state, and the point still holds. Every layer Vercel absorbs is another set of records that lives in Vercel.
A 2026 review scores Vercel’s vendor lock-in at 3 out of 5, or Medium, noting that “zero-config for Next.js = binding to Vercel platform” and flagging that “proprietary edge functions create platform dependency.” It lists Netlify and Cloudflare as the main alternatives and warns that “bills can spike” for cost-sensitive teams.
The pattern is consistent: absorb the layer enough that leaving hurts, keep the marketplace open enough that the lock-in never reads as a trap. That is a pricing strategy, not an accident.
How does this reset the buy-vs-build call?
The decision is no longer “Vercel or Netlify for hosting.” It is whether to accept one vendor for CI, observability, runtime, and agents, and whether to price the switching cost of each layer up front rather than discovering it at renewal.
The second-order consequence is the one platform teams underweight. Every adjacent tool Vercel absorbs is one a competitor can no longer sell against it. A team that picked best-of-breed components (Vercel for hosting, PostHog for analytics, a separate agent framework) now has to justify each external choice against the zero-config bundled version. The bundled version is not necessarily better. It is present, it is on one bill, and it is one fewer vendor to manage. For an understaffed platform team, that is usually enough to win the internal argument.
PitchBook’s competitor set is itself the diagnosis. It lists Netlify among Vercel’s competitors, and a 2026 review names Netlify and Cloudflare as the main alternatives. They are hosts, not platforms. None has acquired the stack above delivery that Vercel has, which is why the buy-vs-build call has shifted toward “buy, from one vendor.”
What can Netlify and Cloudflare still sell?
Delivery and price. The thinner-surface hosts cannot match Vercel’s breadth, but for teams that hit Vercel’s cost ceiling, portability and billing predictability are the wedge that still sells.
The 2026 review warns that “bills can spike” for cost-sensitive teams and names Netlify and Cloudflare as the main alternatives. The structural point is that these competitors contest Vercel on unit economics and exit costs, not on coverage. They can win the team that got burned by a Vercel bill. They cannot win the team that wants analytics, previews, and agents from one vendor without stitching them together.
That is the structural asymmetry the Series D thesis produced. Vercel’s competitors can match it on delivery. To match it on the rest, they would have to buy the same stack Vercel already bought, and pay the integration cost Vercel has already absorbed. The Series D did not literally bankroll every deal; the later Series E and F financed the platform expansion. But the Series D is where the thesis was written, and the thesis is now a structural fact. The platform a team picked in 2021 for fast Next.js deploys is a single-vendor bet on the whole lifecycle, and the remaining question is whether the convenience of one billing line is worth the cost of leaving.
Frequently Asked Questions
Beyond Turborepo and Splitbee, which tools has Vercel pulled into its platform?
The trail continues with Tremor in January 2025 and NuxtLabs in July 2025 on the UI layer, plus three 2026 agent-tooling deals: new.website in March, Hacktron AI in April, and NanoClaw on May 20. PitchBook also lists code-search tool Grep as a Vercel subsidiary, extending the same consolidation into developer search.
Why does PitchBook list Replit, Heroku, and Render as Vercel competitors?
Those three sell backend PaaS and IDE-style surfaces, not static frontends. Their presence in PitchBook’s competitor set signals that Vercel’s stack breadth now overlaps categories beyond delivery, where Netlify and Cloudflare still contest it purely on hosting and unit price.
Which marketplace integration shows a team can keep a layer portable?
PostHog maintains an active Vercel Marketplace integration with over 1,000 installs, covering feature flags, A/B testing, and analytics. For a team auditing which layers are truly bundled versus repointable, that integration is the working example of a feature-set you could move off Vercel without rebuilding, unlike historical data captured at the infrastructure layer.
How solid is the claim that Vercel absorbed a code-search tool?
PitchBook lists Grep as a Vercel subsidiary at medium confidence, but Vercel has not published acquisition terms. Treat the code-search layer as directional rather than confirmed until a vendor announcement or a move inside the Vercel GitHub org corroborates the deal.
What runtime change made long-running agents billable on Vercel?
Fluid, introduced in 2025, let a single instance handle multiple concurrent requests the way a long-running server process does, while keeping serverless elasticity. That shift away from one-function-per-request is what makes an agent that runs for hours a workload Vercel can price, rather than a model that breaks on the first long step.
