- Niteshift, founded by two early Datadog engineers, came out of stealth on June 10, 2026 with a $7 million seed round led by Greylock’s Jerry Chen, plus angels including Reid Hoffman and Datadog’s Olivier Pomel.
- The thesis: companies won’t keep routing their most sensitive code through model makers like OpenAI and Anthropic, which keep launching apps that compete with their own customers — a dynamic some call the “SaaSpocalypse.”
- Instead of selling tokens, Niteshift sells cloud-style infrastructure priced per minute that routes between Claude, GPT, and open-source models so customers avoid single-vendor lock-in.
- It enters a crowded, well-funded field: Cursor, Cognition ($1B raise at a $26B valuation), Amazon Bedrock, and OpenRouter ($113M at a $1.3B valuation) all chase model independence.
“Everybody’s worried about getting stepped on by these giants.” That is how Sajid Mehmood, a former early Datadog engineer, describes the anxiety quietly spreading through teams that have wired their products into OpenAI’s and Anthropic’s coding agents. On June 10, 2026, Mehmood and co-founder Conor Branagan came out of stealth with Niteshift and a $7 million seed round — and a contrarian bet that the safest place for your source code is not inside the company that might compete with you next quarter.
The Bet: Don’t Hand Your Code to a Future Rival
Niteshift’s pitch starts with a question that has become uncomfortable for a lot of engineering leaders. Why would any company route its most sensitive asset — the code that runs its products — directly through model makers like OpenAI and Anthropic, when those same labs keep shipping apps that compete with their own customers? Mehmood has seen this movie before. At Datadog, which he and Branagan helped grow from its early days to a multi-billion-dollar valuation, he watched the monitoring company win e-commerce customers who flatly refused to build on Amazon Web Services — because Amazon was simultaneously putting many of those same retailers out of business during the so-called “retail apocalypse.”
“At Datadog we saw this clearly,” Mehmood told TechCrunch. “A big part of our multicloud business came from e-commerce businesses who did not want to run on Amazon, right? … We are absolutely going to see the same dynamic as Anthropic goes to compete in legal and healthcare and finance and whatever else.” The frontier labs climbing up the stack into vertical software is exactly the shift some founders have nicknamed the “SaaSpocalypse.”
Trend Insight — The lock-in worry stopped being theoretical this month. When the U.S. government abruptly ordered Anthropic to disable overseas access to its top-tier Fable 5 and Mythos 5 models, customers built on those endpoints lost them effectively overnight — a vivid reminder that a proprietary model you don’t control can disappear for reasons that have nothing to do with you.
Selling Infrastructure, Not Intelligence
A coding cloud that routes between models
Crucially, Niteshift is not trying to replace Claude Code or Codex, today’s two most popular coding agents. It argues that it reduces dependence on them. The product is an “AI coding cloud” that routes between models — Claude, GPT, open-source options, and others — based on what each project actually needs. “Being able to switch between GPT and Claude models is important,” Mehmood said. The goal is to separate the coding model from all the orchestration needed to vet, test, and maintain AI-generated code, so the model becomes a swappable component rather than a dependency.
Priced like a cloud provider, not a token meter
The business model is the other tell. Niteshift is not selling tokens; it sells infrastructure and charges like a cloud provider, with per-minute usage rates. “Everybody else is selling labor replacement intelligence,” Mehmood said. “We’re selling software to agents, as opposed to humans — but we’re still out here selling software.” That framing — software for agents rather than a smarter chatbot for humans — is what convinced Greylock’s Jerry Chen to lead the round. “As the frontier labs move up the stack, there’s an opportunity to offer customers an alternate path: unbundling their agents from the infrastructure they run on,” Chen said. “Niteshift is building the platform that enables this for coding agents, letting customers invest deeply in their developer tooling without locking themselves into a single model or agent vendor.”
Trend Insight — The seed round is small at $7 million, but the cap table is a statement: Reid Hoffman, Datadog’s Olivier Pomel and Alexis Lê-Quôc, Braintrust’s Ankur Goyal, and Reflection AI’s Misha Laskin all bought in. When operators who built the last generation of developer infrastructure back “model independence,” it signals where they think the durable value will sit.
The Bigger Trend: AI’s Unbundling Moment
Niteshift is not alone, and that is the real story. Model-agnostic routing has suddenly become a crowded, well-funded category. AI gateway OpenRouter more than doubled its valuation to $1.3 billion on a $113 million raise; Cognition pulled in $1 billion at a $26 billion valuation; Amazon Bedrock pushes multi-model access from the cloud side; and Cursor, the breakout coding tool, is reportedly being circled by SpaceX. At the same time, the industry is openly debating whether companies can learn to love cheaper, smaller models instead of always reaching for the priciest frontier system, while an “open-source AI must win” manifesto circulates among developers who want AI treated as shared infrastructure rather than something gated behind a handful of vendors.
Trend Insight — The throughline is unbundling. The first phase of the AI boom rewarded whoever had the single best model. The next phase may reward whoever lets you treat models as interchangeable parts — swapping Claude for GPT for an open-weight model the way cloud teams already swap regions and instances.
What It Means for Builders
For engineering leaders, the practical takeaway is to design for portability now, before switching costs harden. In practice that means keeping an abstraction layer between your application and any single model API, asking whether your most sensitive workloads should really run through a vendor that also sells into your market, and treating “Can we move off this model in a week?” as a genuine architectural requirement rather than a hypothetical. Niteshift may or may not be the company that ultimately wins this category — it is entering late against well-capitalized rivals — but the question it is forcing onto the table is one every team shipping AI features should be able to answer.
Related
- A 5:21 PM Letter Just Took Down Two AI Models
- Europe’s $14B AI Bet Is Coming for the Factory Floor
- Perplexity Let AI Write Its Own Search — And It Won
- This AI Quietly Captured 40% of the Fortune 50
- Why Meta’s Next AI Hub Isn’t in America
Sources
- TechCrunch — Datadog veterans launch AI coding startup Niteshift on a bet against Big AI lock-in (Julie Bort, June 10, 2026)
- TechCrunch — Can tech companies learn to love cheaper AI models? (Russell Brandom)
- GeekNews — Open-source AI and model-switching discussion (news.hada.io)
AI Biz Insider · AI Trends EN · aibizinsider.com
댓글 남기기