By Mollie Barnett
When Andrej Karpathy joined Anthropic this week, most people in tech treated it like another AI talent headline.
It is not.
Karpathy is one of the most respected minds in artificial intelligence. Founding member of OpenAI. Former head of AI vision at Tesla. Stanford educator. Builder of nanoGPT and micrograd. One of the clearest AI educators alive.
He could have gone almost anywhere.
Meta. xAI. Google DeepMind. OpenAI. His own venture, full time.
He chose Anthropic.
That matters because the AI industry is no longer just competing on model capability. It is competing on trust, control, infrastructure, and moral legitimacy.
And that shift does not stop at Silicon Valley. It reaches Long Island businesses too.
The major AI players are beginning to separate into distinct philosophies. OpenAI increasingly behaves like a massive commercial platform: distribution first, ecosystem scale close behind.
Google DeepMind remains tied to one of the most powerful centralized information systems in the world: search, productivity software, cloud infrastructure, chips, and data.
Anthropic has positioned itself around safety, interpretability, and responsible scaling.
Musk’s xAI frames itself around transparency, anti-censorship, and resistance to centralized narrative control.
These are not just brand differences. They are different theories of power.
Who should build AI?
Who should govern it?
Who should own the infrastructure?
Who should control the information layer?
Who gets access?
Who gets priced out?
That is the real split emerging underneath the product launches. There is also a question forming around OpenAI that deserves more attention.
As OpenAI deepens its alignment with federal contracts and defense infrastructure, the category itself starts to divide structurally. If one major model becomes deeply embedded in government and military operations, then every other model becomes an alternative by definition.
That has consequences for public trust, procurement, business adoption, and what “independent AI” even means.
These are not abstract questions. They are arriving quickly.
Now zoom out further.
Google has more AI infrastructure than almost anyone in this race: DeepMind, Gemini, search, Workspace, Cloud, YouTube, Android, and the chips that power its models while much of the market remains dependent on Nvidia.
Blackstone just announced a major joint venture with Google to build and operate AI data centers using Google’s own chips. Berkshire Hathaway also sharply increased its Alphabet position in Q1 2026.
Whether coincidence or convergence, the signal is hard to ignore.
Google is not just competing for AI market share. It may be positioning around the infrastructure layer that other markets eventually depend on. That matters because the future of AI will not be decided only by which chatbot feels smartest.
It will be decided by who owns the compute, who controls the distribution, who sets the defaults, who captures the data, and who becomes too embedded to avoid.
And then there is China.
DeepSeek exploded into public attention by offering powerful AI at little or no cost. Kimi and other Chinese AI systems are also gaining serious enterprise and consumer traction.
The function is real. The price is low. But nothing is free.
When powerful AI tools are free, the strategic question is not just, “Does this work?”
It is: Where does the data go? Who can access it? What legal regime governs it? What happens when millions of users and businesses trade cost savings for dependency?
This is where American AI has a serious vulnerability. If U.S. AI remains priced primarily for enterprise buyers while Chinese AI remains free or nearly free, many small businesses, independent workers, students, and local organizations will make a rational choice.
They will use the free tool, not because they are naïve, because payroll is due. Rent is due, margins are thin, and access matters.
That is how America could lose the data race without losing the technology race. We could build the most capable, most safety-conscious, most values-aligned AI systems in the world — and still price our own people out of using them.
That is not just a market problem. It is a sovereignty problem, and it is why local businesses need to start thinking differently about AI adoption.
Most companies are still using AI as a tool: a faster way to write, summarize, automate, or finish tasks they were already doing. That is useful, but it is not the real leverage. The bigger opportunity is orchestration.
Automation replaces a task. Orchestration rewires how intelligence moves through an organization. It changes how decisions get made, how workflows connect, how knowledge compounds, how teams coordinate, and how leaders see around corners.
That is where the tenfold returns begin.
The practical response is not panic. It is fluency. Business leaders do not need to become machine learning engineers, but they do need enough AI literacy to understand what they are adopting, what they are exposing, what they are depending on, and where leverage actually lives.
AI adoption is becoming a sovereignty decision, not only for nations, for companies, for communities, for schools, for local economies, for Long Island businesses deciding which systems will shape how they work, compete, and grow.
The businesses that learn orchestration now will move faster, decide better, and outcompete organizations ten times their size still running on legacy logic.
The ones that do not may not disappear overnight. They may simply wake up one day operating inside someone else’s platform, someone else’s defaults, and someone else’s economic model.
That is the shift. AI is no longer just a productivity tool.
It is becoming infrastructure, and infrastructure always becomes power.
