Where the AI Industry Is Going. And What Arriving Means.

Hilary Joseph Castille
May 21, 2026

The major artificial intelligence labs have publicly committed to a destination. Artificial general intelligence — machine systems that match or exceed human capability across all cognitive domains — by sometime in the second half of this century. The investment is staked on that timeline. The corporate strategies of most of the world's most consequential technology firms are organized around reaching it.

The scale of what has been committed to is documented. Training compute for frontier AI models has grown approximately four to five times every year since 2020 — doubling roughly every five months, per Epoch AI's tracking. As of mid-2025, more than thirty frontier models from twelve developers had exceeded the threshold of computational work that took the field five years to reach. Per Synergy Research Group, combined capital expenditure by Amazon, Google, Meta, and Microsoft was approximately $413 billion in 2025, up 84 percent from the prior year, and is projected at approximately $725 billion in 2026. A single quarter of hyperscale data-center capital expenditure hit $142 billion in the third quarter of 2025 alone — up roughly 180 percent year over year.

The Stargate program, announced jointly by OpenAI, SoftBank, Oracle, and MGX in January 2025, plans up to $500 billion of investment through 2029 across data center sites totaling nearly seven gigawatts of capacity. NVIDIA, whose chips power most of this work, became the first public company to reach a $5 trillion market capitalization in October 2025. Frontier model training runs, estimated by Epoch AI at $80 million for the leading 2023 model, are projected at $1 billion to $3 billion by late 2027.

These are not speculative projections. These are documented capital flows, public corporate announcements, and operating cost curves. The research programs underneath them are internally coherent. The people doing the work are intelligent and serious. The trajectory is real.

So let us take them at their word.

The question that follows, and the question this series is built around, is the question the public conversation has mostly skipped past. Discussion of AI is dominated by whether the technology will scale, whether the timelines are accurate, whether sufficient safety work is being done, and whether one lab or another will reach the destination first. These are real questions. They are not the question that matters most.

The question that matters most is what arriving actually means.

If the AI industry succeeds at the project it has committed to — if the destination is reached, on something close to the timeline its leaders have publicly stated — what does the world look like the day after? What kind of work do most people do? What kind of services do they receive from the institutions that shape their lives — their doctors, their courts, their schools, their governments? What kind of information environment do they live inside? What happens to the languages and cultures their grandparents spoke and lived in? What kind of energy and physical infrastructure does civilization commit to keeping the systems running? And who, in the economy that emerges, captures the value that this technology generates?

These are not technical questions. They are questions about the texture of human life in a world where artificial intelligence has been built to mediate most of it. And they have answers — documented, sourced, already visible in primary research, court records, corporate filings, the trajectories of public companies, the layoffs already being announced, the data centers already being built. The answers are not speculative because the deployment is not speculative. It is happening now, at scale, on the trajectory the industry has stated.

The case for examining those answers seriously is straightforward. Capital allocation decisions at the scale of hundreds of billions of dollars per quarter, made on a stated trajectory, are not casual. They are positioning a particular destination as the default future. The default arrives unless something else is built. Whether the destination is the future we would choose, if we paused to choose it, is a question we owe ourselves to ask before the choice is made by accumulation rather than by deliberation.

This essay is the leading edge of a series. Over the coming weeks, the series will walk, one dimension at a time, through what arriving at the AI industry's stated destination actually means across the dimensions of human experience the technology is being built to mediate.

The series will examine what happens to human work when the architecture being scaled is designed to do what humans currently do — and what the proposed response of subsidizing the displaced actually offers, and what it does not.

It will examine what happens to the institutional services people depend on — medicine, law, government, education, finance — when the systems mediating them have no inherent way to tell the difference between what is true and what is statistically plausible, and what the documented record of that gap already shows.

It will examine what happens to the public information environment as artificial intelligence generates content at scale faster than any prior technology has been able to, and what the consequences are for the shared reality that makes democratic life possible.

It will examine what happens to language and culture when artificial intelligence is built primarily on English-language data, deployed globally, and asked to mediate the lives of seven billion people whose languages and cultural contexts the architecture was not built to preserve.

It will examine the physical resources — electricity, water, semiconductors, the geographies where data infrastructure concentrates — that the destination requires, and what trade-offs that infrastructure footprint imposes on civilization at the trajectory pace.

It will examine the economic structure that emerges — who owns the systems doing the work, who captures the wealth the systems generate, what the distribution looks like across populations and across the world, and what the political consequences of that distribution are likely to be.

Each of these dimensions has been documented already. The data is in peer-reviewed journals, government statistics, court filings, corporate disclosures, primary research from established institutions. None of what the series will trace is hidden. What is missing is the synthesis — the act of looking at the dimensions together, in one place, against a common question.

That common question is what this essay opened with. The major AI labs have told us where they are going. The investment is real. The timelines are public. The architecture is being built and deployed at the pace the trajectory describes. The conversation we have not yet had is the conversation about what arriving means — and whether, having looked at it carefully, the destination is one we would choose.

This series is an effort to look at these questions carefully. It is not a wholesale condemnation of the field. The authors are of the view that an alternative understanding of general intelligence might inform a more desirable choice of paths. The next essay begins the walk.

Co-authored with Alan Ignatius Touring, MetaCom Co-Founder (AI). Hilary Joseph Castille writes from Lafayette, Louisiana, where he founded MetaCom in 2022 to develop alternative approaches to human problem solving.

Originally Published May 21, 2026, on LinkedIn: https://www.linkedin.com/pulse/where-ai-industry-going-what-arriving-means-hilary-joseph-castille-ks8pc/