#vendor-lock-in

#vendor-lock-in

Last week at AWS re:Invent, amid many product announcements and cloud messages, AWS introduced AWS AI Factories. The press release emphasizes accelerating artificial intelligence development with Trainium, Nvidia GPUs, and reliable, secure infrastructure, all delivered with the ease, security, and sophistication you've come to expect from Amazon's cloud. If you're an enterprise leader with a budget and a mandate to "do more with AI," the announcement is likely to prompt C-suite inquiries about deploying your own factory.

Amazon's Nova 2 announcement at AWS re:Invent 2025 is exactly the type of AI offering we expected from AWS and, frankly, exactly what should make thoughtful architects nervous. Nova 2 is positioned as a frontier-grade model, tightly integrated with Amazon Bedrock. It's part of a growing ecosystem of "frontier agents" and the AgentCore framework unveiled at re:Invent 2025. The story is compelling: better models, better tools, and a single platform to build, deploy, and scale agentic AI.

The commitment follows a string of setbacks in the government's handling of the largest tech suppliers, especially in the provision of cloud services. In February 2024, The Register reported that officials from the from the Central Digital & Data Office admitted the government's approach to cloud adoption had resulted in: "risk concentration and vendor lock-in that inhibit UK government's negotiating power over the cloud vendors."

Consider the U.S. Department of Veterans Affairs, which recently committed $4.7 billion to Microsoft over five years, probably to repurchase software it has used for decades. This new contract obligates $940 million a year, almost double their previous $1.6 billion, three-year agreement. That's about $27 for every tax-filer in America. It also means the VA's spending on Microsoft has nearly doubled in just three years. Multiply this across many agencies and different vendors and the true cost to taxpayers becomes staggering.

Given that corporate IT relies heavily on cloud-based infrastructure and services delivered via the public cloud, access to the data held in the cloud is paramount. Should all mission-critical data be held on-premise? What roles should digital sovereignty and digital residency play in a corporate IT strategy? These are among the questions being discussed at Forrester's forthcoming Technology & Innovation Summit in London.

Shortly after Reuters broke the news that Meta, the company behind Facebook and Instagram, had signed a $10 billion, multiyear cloud deal with Google, a CIO friend called me with some questions: "David, should we be doing something like this? Are these mega prepurchase cloud agreements the new table stakes for enterprises?" The question is as pressing as it is complex. My answer, shaped by years of watching and guiding digital transformation, may not be what cloud vendors want you to hear.

The choice between Linux and MacOS isn't hard. If you can answer these questions, you'll know which to choose. Both are outstanding choices and will serve you well. I use both Linux and MacOS. The former is used for everyday tasks, and the latter for video editing and mobile usage (please, someone, create a Linux laptop that is as reliable and similar to a MacBook).

Not to be outdone by the makers of ChatGPT and Claude, who each agreed to sell their services to the government for $1 per agency, Google has agreed to even deeper discount terms, pitching its various government-capable AI products for just $0.47 per agency, valid through 2026. The half-a-buck Google AI deal is part of the General Services Administration's OneGov purchasing strategy that seeks to streamline the purchasing of products for federal agencies.