The core lesson Sajid Mehmood took away from Datadog was simple. "Customers do not hand their core infrastructure over to their competitors."
That pattern is playing out again in the AI coding tool market right now. As OpenAI and Anthropic openly declare their plans to enter vertical markets, if you are relying on their AI as your core development infrastructure — it is natural to start feeling uneasy about that structure.
Datadog did not beat AWS with better technology
AWS CloudWatch was already in the cloud monitoring market. It was free and perfectly integrated with AWS infrastructure. Yet independent startup Datadog took over the market.
Mehmood witnessed the reason firsthand. Many of Datadog's customers were companies competing in the same markets as Amazon — retail, logistics, media. They did not want their infrastructure data going to a competitor. Regardless of tool performance, relying on a competitor platform felt strategically uncomfortable.
What is Supplier Conflict?
A structural conflict of interest that occurs when the company behind the tools you use also competes in your market. Beyond functional merit, it creates strategic risk — your core infrastructure data flowing to a competitive party.
Notable in this context: Datadog co-founders Olivier Pomel and Alexis Lê-Quôc directly invested in Niteshift. They know better than anyone that the pattern they created is repeating itself in the AI market.
The same structure is emerging in AI coding
Here is Mehmood in his own words: "At Datadog we saw this clearly. We are absolutely going to see the same dynamic as Anthropic goes to compete in legal and healthcare and finance and whatever else."
To put the scale of the AI coding agent market in context — Cognition (makers of Devin) raised $1B at a $26B valuation, while multi-model routing layer OpenRouter raised $113M and now serves 25 trillion tokens per week to 8 million developers across 400+ models.
From the OpenRouter Series B announcement: As organizations move from single-model pilots to multi-model production systems, they need a routing and gateway layer purpose-built for that complexity.
"Being able to switch between GPT and Claude models is important. Everybody is worried about getting stepped on by these giants."
— Sajid Mehmood, Niteshift CEO
Niteshift is not pitching a simple model-switching tool. It is cloud infrastructure for running AI coding agents — billed per minute, not per token. As Mehmood puts it: "Everybody else is selling labor replacement intelligence. We are selling software to agents, as opposed to humans."
Things to check right now
Even while Niteshift is still early stage, the practical lessons from this trend can be applied today.
- Understand your vendor expansion plans
Find out whether your AI coding tool model provider (OpenAI, Anthropic, etc.) plans to launch products directly in your industry. If you are in a market where they could become a competitor, reducing single-vendor dependency is the strategic move. - Review data handling policies
Check the ToS of your current AI coding tools for data training clauses. Some, like Cursor Enterprise, explicitly specify a Privacy Mode — no training on your code — while others do not. Beyond training, knowing where your data goes is basic due diligence. - Explore multi-model routing layers
Get familiar with options like OpenRouter (400+ models, API integration), Amazon Bedrock, and Niteshift. Even if you do not switch today, keeping your options open is itself a form of risk management.
| Single-vendor dependency | Multi-model routing | |
|---|---|---|
| Supplier conflict risk | High — vulnerable if they enter your market | Low — can switch vendors anytime |
| Cost optimization | Locked to one model pricing policy | Auto-select best model per task type |
| Model performance changes | Directly exposed to provider updates | Auto-route to better performing models |
| Initial setup complexity | Low | Medium (routing layer configuration needed) |
Greylock Jerry Chen, Reid Hoffman, and Datadog founders betting $7M on Niteshift reflects conviction that this pattern repeats. The way Datadog beat AWS will play out again in AI infrastructure.
More to explore
Datadog Veterans Launch AI Coding Startup Niteshift The full TechCrunch article with Mehmood background and founding story in detail. techcrunch.com
OpenRouter Series B announcement Best snapshot of the multi-model routing market scale and direction. CapitalG and NVIDIA backing, 400+ models, 25T tokens/week. openrouter.ai
Cursor Enterprise Where to verify Privacy Mode and current multi-model support (OpenAI, Anthropic, Gemini, xAI) in official documentation. cursor.com
Claude for Enterprise Anthropic official position on enterprise market definition and data policy. claude.com
Amazon Bedrock AWS official multi-model platform — Claude, Llama, Stable Diffusion and more through a single API. Battle-tested enterprise option available today. aws.amazon.com
