The Business Engineer

Premium Access to PDF of Book

Part One

The Abstraction — A Second Computing Revolution

The mistake most analysts make is treating AI as a continuation of the web. It is not. The web was a distribution revolution layered on top of an existing computer. AI is the computer being rebuilt.

That single sentence reframes everything. It tells you why CapEx in 2026 will cross one trillion dollars and still leave the demand curve underserved. It tells you why memory contracts are sold out through 2027. It tells you why hyperscalers are signing nuclear power deals. It tells you why every frontier lab is also, quietly, becoming a chip company. The web revolution took a decade. The computing revolution that produced the web took fifty years. AI belongs to the second category.

The cleanest mental model I have for what is happening: the agent is becoming the computer. Until very recently your laptop was a deterministic machine — you typed a command, it executed deterministically. Right now, on millions of machines, an AI agent is sitting at the terminal, sandboxed, controlling the operating system through MCP, executing probabilistically on your behalf. In two or three years, the terminal-plus-agent collapses into a single object. The computer becomes probabilistic. The operator becomes the supervisor.

If you accept that frame, every other piece of the analysis falls into place. The trillion-dollar CapEx is not a corporate spending cycle — it is the cost of rebuilding the substrate. The geopolitical race is not over markets — it is over which civilization owns the new computer. The “bubble” question is not whether demand is real — demand is unconstrained — but whether physics can deliver supply fast enough.

There is also a historical rhyme worth holding in mind. In 1957, eight researchers left Shockley Semiconductor and founded Fairchild — the Traitorous Eight. Their exit gave rise to Intel, the venture capital model, and the entire Silicon Valley. In 2021, eight researchers left OpenAI and founded Anthropic. We are watching the same pattern at the same point in the cycle, with the same long compounding ahead.

The Market Map — Nine Layers, Three Geometries, One Backbone

The Nine Layers

Every layer has a bottleneck. Not a soft bottleneck — a physical one. HBM4 is sold out through the second half of 2027 across the three suppliers that produce it (SK Hynix, Samsung, Micron). CoWoS advanced packaging at TSMC is allocated to NVIDIA first; merchant silicon teams negotiate for scraps. The U.S. grid is not designed for AI; only China’s grid is, because China invested in nuclear for non-AI reasons and now harvests the optionality. Lithography sits behind a single Dutch company.

The cleanest reading of the map is this: demand is unconstrained, supply is constrained everywhere. That is the inverse of what a classical bubble looks like. In a bubble, supply races ahead of demand and capital chases imaginary returns. In this cycle, capital is hitting a wall of physics. The bubble risk is not over-investment in compute — it is misallocation across layers when one bottleneck migrates to another faster than balance sheets can adjust.

The Trillion-Dollar Year

The 2026 CapEx number is roughly $1.04 trillion, distributed as follows: Big Four (Amazon, Alphabet, Microsoft, Meta) at $725B, China at $80B, neoclouds at $60B, sovereigns at $60B, Oracle at $50B, Apple at $13B, and the long tail at $48B. The Big Four alone are 3.7× their combined 2023 spend.

My base case is that 2026 closes well above the trillion-dollar mark — call it $1.10–1.15T — because two wildcards have not yet ramped. Apple is ending its decade-long net-cash-neutral posture; the new CEO will be forced to spend, because Apple is no longer TSMC’s largest customer (NVIDIA is) and the constraint is now on Apple’s ability to produce its own devices, not on its capital. SpaceX is the second wildcard — Colossus, the 220K-GPU cluster, plus a vertical-integration thesis that extends from launch (90%+ market share) to satellite (Starlink, 4M+ subscribers) to compute to, in the S-1, an in-house fab.

What does a trillion dollars buy in 2026? Roughly 75 to 90 gigawatts of new power capacity by 2028 — equivalent to the entire UK grid. Roughly 15 to 18 million GPU-equivalent units. A multiple of current HBM production. It buys enough supply to serve the next two to three years of demand, assuming no new demand surprise. The five-year cumulative need, on current trajectories, is closer to five to six trillion dollars.

Three Geometries (Plus a Fourth)

There are three solutions to the problem of “how do I capture rent across a layered stack” — plus a fourth that the niche players have found.

Vertical (Google). Own a column from foundry partner to user. Google has the TPU (sixth generation, deployed at scale, now sold externally for the first time in 2026), the XLA software stack, the Gemini model family, the agentic harness inside Gemini, and the distribution surface in Search plus Android. Every cost reduction at any layer flows through the entire column. The disadvantage: vertical columns are consumer-first by construction, and the enterprise battle requires different muscle.

Horizontal (NVIDIA). Own a band across all customers. Silicon (Blackwell, Vera Rubin), networking (CUDA, NVLink, Spectrum-X, Mellanox), systems (DGX, NVL72), inference software (NEMO, NIM), and an enterprise harness layer (AI Enterprise, Omniverse). NVIDIA is not a chip company — it is the connective tissue. Every GPU dollar drags a networking dollar; that ratio approaches parity at 100,000-GPU scale. The Mellanox acquisition is, in hindsight, the highest-return semiconductor deal of the past decade.

Flywheel (SpaceX). Multiple reinforcing engines that each feed the others. Launch capacity funds satellite buildout; satellites fund compute; compute funds the next launch. SpaceX does not fit the vertical or horizontal templates because the engines are not in the same stack. They are in adjacent stacks, and the cross-feed is the moat. This is the wildcard that made me redraw the map.

Niche by Adjacency. Anthropic owns foundation models plus agentic harness, deeply. Databricks owns compute capacity plus models for data-platform use cases. CrowdStrike owns harness plus distribution on the endpoint. The lesson is durable: you do not need nine layers — you need two adjacent ones, owned completely.

The Whale Backlog

The remaining performance obligations on the cloud books tell you where the demand is parked. Combined backlog across Microsoft ($627B), Oracle ($553B), Alphabet ($468B), and Amazon ($464B) is roughly $2.1 trillion. Of that, somewhere on the order of half traces back to two counterparties: OpenAI and Anthropic.

That circularity — hyperscalers invest equity in labs, labs commit compute credits back to those hyperscalers, hyperscalers book the revenue — is real. The honest read is that it works as long as the labs grow into the obligations. The required rate of growth: 20–30× over three years. OpenAI from ~$5B ARR today to $100–150B. Anthropic from ~$9B (some say $30B+ projecting forward) to $180–270B. No software company in history has done 20× in three years. Both labs need to do it simultaneously.

That is what makes the upcoming IPOs (SpaceX, OpenAI, Anthropic) the most important test of the entire cycle. The private market can carry the assumption. The public market will price it.

The China Parallel Stack

Export controls did not slow China — they forced China to build an indigenous full-stack from silicon to distribution. Huawei Ascend on SMIC 7-nm fabrication, CXMT and YMTC memory, MindSpore and CANN as the software layer, DeepSeek and Baidu and Alibaba at the model layer, domestic distribution platforms on top. The capability gap to the U.S. frontier is three to seven months. The sovereignty gap is zero — full end-to-end independence.

This is why the governance layer matters geopolitically. When Anthropic and OpenAI release their most capable models on a paced, capability-gated cadence, the side effect — possibly the intent — is that Chinese open-source labs lose their distillation source. Open-weight Chinese models stayed near the frontier in part because they could distill from the U.S. frontier through API. Pace the frontier release, and you cap the open-source ceiling.

The Playbook — How to Read the Map

Cascades, Not Stacks

The stack metaphor is useful for inventory but misleading for dynamics. The real action is in cascades — how a shift in one layer propagates to every adjacent block.

Take the most important cascade flowing downward right now, from harness to silicon. Agentic queries consume roughly 500× the tokens of single-turn chat queries because every agent call chains multiple model invocations. That makes cost-per-token a first-order constraint. The economics of inference at agentic scale demand cheaper inference at massive volumes. That pressure pushes open-weight models to co-design with silicon. Inference-specialized clouds emerge. Silicon stops being a component and becomes a service, priced by the query rather than by the wafer.

Or take the cascade flowing upward through the financing stack. NVIDIA leverages GPU allocation plus financing to lock in AI labs. AI labs consume compute and ramp up revenue commitments back to the clouds. Clouds book backlog. Backlog supports more capex. If labs fail to monetize, cloud backlog is impaired, NVIDIA demand drops, the cycle reverses. Every dollar in the system is someone else’s revenue and someone else’s cost.

Two Structural Breaks

Two events in the past twelve months were unthinkable eighteen months ago.

Microsoft–OpenAI exclusivity ends. OpenAI can now serve from any cloud. Azure loses its only moat in model serving — the $627B Microsoft backlog is suddenly exposed to multi-cloud risk. Microsoft’s response: ramp CapEx urgently to lock down compute, accelerate the development of its own frontier model in-house, and try to make Copilot the GitHub-of-AI-coding before Claude Code finishes eating that lunch. The honest read on the supposed Microsoft cancellation of Claude Code licenses: it is not about token cost — it is about not letting a competing harness eat your enterprise developer relationship.

Apple ends net-cash-neutral. For the first time in the modern era, Apple will hold more debt than cash. CapEx intensity actually declines as a share of revenue (5% to 3.5%), but absolute spend rises. The strategic problem for Apple is not financial — Apple has plenty of cash. The strategic problem is that NVIDIA, not Apple, is now TSMC’s largest customer. Apple has lost the pace-setting position it held for fifteen years. In a near future where consumers want multiple devices to run multiple local agents, Apple may not be able to produce enough chips to meet its own demand. That is the supply-side risk Tim Cook’s successor was hired to fix.

Hyperscalers Become Utilities

The bottom of the stack used to be compute. Now it is substations. Microsoft is restarting Three Mile Island for 835 MW. Amazon has signed Talen for nuclear plus gas portfolios totaling 1,920 MW. Meta is building Project Hyperion, a 2-GW single-campus data center, the largest ever. Google is investing in Oklo small modular reactors. These are not data center deals — they are utility deals. A hyperscaler in 2026 is, structurally, a power company that happens to sell tokens.

The corollary: power geography becomes destiny. Pennsylvania for nuclear-plus-coal baseload. Texas, the Nordics, and Quebec for cheap stranded hydro and nuclear. The Gulf states for unlimited solar plus gas, backed by sovereign wealth — Gulf SWFs collectively hold roughly $2.5T and are pivoting hard from oil to AI infrastructure. NEOM’s smart-city budget was cut 58% and redirected to AI compute. Self-generation through on-site SMRs is the wildcard for those who cannot wait for grid permitting.

The Memory Crunch

If you want a single price signal that tells you the cycle is real, look at HBM contract prices. They moved 95% quarter-over-quarter. The oligopoly is three players and the production is allocated through 2027 to a single buyer with priority. Every consumer-electronics company in the world will pay a DRAM premium as collateral damage. This is what supply-bound demand looks like in a chart.

The memory crunch will persist for years because semiconductor capacity does not scale on annual cadence — it scales on five-to-ten-year cadence. The trillion dollars of 2026 CapEx buys you the foundation for 2027–2028 supply. To serve 2029–2030, you need another trillion in 2027.

Capital Equals Physics

The simplest mental model for this market: capital equals physics. As long as physics has slack, capital compounds. The moment physics binds — and it binds now, at every layer — capital stops being the variable. Permitting becomes the variable. Lithography slots become the variable. Substrate yields become the variable.

The IPO window is the public test of how this rebalances. SpaceX, OpenAI, and Anthropic going public in the next eighteen months will tell us whether the market accepts the capex-equals-physics frame and prices accordingly, or whether the public markets impose a different rate of return on the labs than the private markets have. My base case is the former. The bear case is the latter, and the bear case is real.

What’s Next — The 2027–2028 Inflection

A few patterns I expect to harden over the next eighteen to twenty-four months.

The four scaling paradigms collapse into recursive learning. Right now, four scaling laws stack: pre-training scaling, post-training reasoning, test-time compute, and agentic loop scaling. They run in parallel silos. The next breakthrough — and Andrej Karpathy joining Anthropic is the signal — is recursive self-update: models that update their own weights from agentic feedback. If that lands, you get continual learning across all four scaling laws simultaneously. That is the architecture that pushes us past the current quality plateau toward something most people would recognize as AGI.

Agentic enterprise becomes the default. Two vendors, one protocol layer. Claude for reasoning, Codex for automation, both interoperating through MCP and AGENTS.md. MCP plus AGENTS.md are becoming the HTTP-and-HTML of agents. Proprietary walled gardens are losing; standards are winning. Every enterprise will run a two-vendor harness by end of 2027.

Distribution becomes the consumer endgame. Apple is embedding Gemini into 2.2 billion devices. Google AI is becoming Google search for billions of users who will never know it changed. Meta needs the AR glasses bet to land because that is where the consumer agent sits on the body, not on the desk. ChatGPT consumer share peaks; super-app dynamics take over.

The China stack completes. By 2027, every layer of the Chinese AI economy will be domestic, with capability roughly six months behind the U.S. frontier and full sovereignty. The strategic question is no longer whether the bifurcation happens — it has happened — but whether the two stacks remain interoperable at the protocol layer or fully decouple.

Sovereign AI becomes a permanent third demand category. Beyond hyperscale and enterprise, sovereign compute is now structural. Every nation-state with the means is building national capacity. Every sovereign program buys NVIDIA at the top of the stack because there is no alternative at that scale.

Networking eats economics. At 100K+ GPU clusters, networking cost approaches GPU cost. The Mellanox acquisition becomes the most consequential deal NVIDIA ever made. Every model release will be benchmarked against the networking architecture it requires to serve.

Physical AI emerges as the third value region. Infrastructure (silicon, compute, energy) keeps the highest margins for the longest time. Agentic applications (models, harness, distribution) get the fastest growth and the most competitive intensity. Physical AI (robotics, autonomous, AI-RAN, industrial) is the earliest stage with the largest TAM and a thirty-year deployment horizon. Every humanoid robot and every autonomous vehicle and every AI-RAN telecom defaults to NVIDIA’s reference architecture. The cascade flows.

The TSMC question is the choke-point of the choke-points. TSMC is the constraint that everything else routes through, and it is not ramping CapEx as fast as the rest of the system wants. Read that as discipline, not weakness — TSMC remembers the early-2000s telecom CapEx hangover and is deliberately holding the line. But it means the supply ceiling is set in Taiwan, with all the geopolitical exposure that implies.

Key Takeaways & Mental Models

Five frameworks to carry from this map into the next twelve months of decisions.

Capital = Physics. Capital flows where physics has slack. Capital halts where physics binds. In 2026 every layer of the AI stack has hit a binding physical constraint, so capital’s role flips from accelerator to allocator.
The Agent Becomes the Computer. The probabilistic agent merges with the deterministic operating system within three years. Every product strategy should answer one question: what does this look like when the agent is the computer, not a feature inside it?
Demand is Unconstrained, Supply is Bound Everywhere. The bubble frame is wrong because the cycle is supply-bound, not demand-bound. The real risk is misallocation across the nine layers as the binding constraint migrates faster than balance sheets can adjust.
Two Adjacent Layers Beats Nine. The niche-by-adjacency pattern — Anthropic on Models+Harness, Databricks on Compute+Models, CrowdStrike on Harness+Distribution — is the winning posture for anyone who is not Google, NVIDIA, or SpaceX. Pick two adjacent layers, own them completely, ignore the rest.
Governance is the New Choke Point. Paced release of frontier capability is now a deliberate competitive lever, not just a safety posture. It caps open-source distillation, it differentiates enterprise positioning, and it shapes the geopolitical balance between the two stacks.

Part Two - The AI Capex Dynamics

A trillion dollars in capex this year buys roughly two years of AI supply.

That sentence does almost all the analytical work in this piece. Once it lands, the noise about circular financing, the bubble debate, the off-balance-sheet vehicles, and the IPO scramble all snap into the same picture. We are not in a cloud-capex cycle. We are not in a web-investment cycle. We are in a computer-rebuild cycle, and the physical infrastructure required to rebuild a computer at planetary scale costs more than the public market is yet prepared to price.

Three layers of analysis: the abstraction (why this is a second computing revolution, not a web cycle), the market map (the financing machine that funds the rebuild), the playbook (three capital cascades and what to watch in each), and what comes next (the wildcards that will push 2026 capex past the trillion-dollar mark and the IPO window that will test the whole thesis).

Salesforce and The AGaaS Transition

Gennaro Cuofano — Thu, 28 May 2026 11:59:39 GMT

Every business-model transition produces one company whose results you read not to understand the company, but to locate the entire category.

For the shift from licensed software to the cloud, that company was Salesforce.

For the shift now underway — from Software-as-a-Service, where firms charge for access to tools that humans operate, to Agentic-as-a-Service (AGaaS), where they charge for outcomes that agents execute — the instrument is Salesforce again.

The first quarter of fiscal 2027, reported May 27, 2026, is the cleanest public reading we have of where that transition really sits. And the pattern beneath the headline is consistent and important:

Salesforce has rebuilt its story, its metrics, and even its chart of accounts around AGaaS — while the mechanics underneath remain seat-priced SaaS with consumption bolted on the side. Every gap between the two is a coordinate on the transition map. Plotted together, they tell us the market is at the threshold — not the old world, not yet the new one.

The Interactive AGaaS Defensibility Map

The Portal collapses seven assets I have built separately over the past decade into one substrate.

The AI Fog & The New Judgment Layer

Gennaro Cuofano — Wed, 27 May 2026 04:59:47 GMT

Two structural shifts — AGaaS and Agent Expansion — are converging on a single conclusion: the human is being evicted from the interface and execution layers of work, into a single remaining layer that has no name yet, no curriculum, and no playbook.

Call it the judgment layer. The fog is what it feels like from inside.

Request Access To HQ Full Map Here!

The AI Economy Workshop

Gennaro Cuofano — Tue, 26 May 2026 11:59:49 GMT

A trillion dollars in capex this year buys roughly two years of AI supply.

For Premium Members:

Get The Business Engineer Agent Here!

Try The Business Engineer Agent Here

The Agentic Expansion Cascade

Gennaro Cuofano — Mon, 25 May 2026 04:59:53 GMT

The agentic expansion theory — sandbox → tool use → computer use → convergence — is usually read as a product roadmap. Each lab ships the next phase; each phase looks like a feature release.

The structural reading is different. The expansion is not a product decision. It is the substrate requirement of the scaling law that now defines AI progress. Once that is seen, the cascade across the rest of the AI map is not a forecast — it is a settlement.

This piece has two parts. Part I explains why the expansion is happening, from the AI scaling perspective. Part II runs the cascade forward across the seven layers of the AI map that result in a reprice.

PART I — Why This Is Happening

The Scaling Laws Stack. They Do Not Sequence.

The standard narrative treats AI scaling as a sequence. Pre-training was 2022–2023. Post-training was 2023–2024. Test-time compute was 2024–2025. Agentic loops are 2025–2026. Each paradigm “replaced” the last. That framing is wrong in a way that matters for capital allocation and product strategy.

None of these paradigms got replaced. Pre-training is still compounding — Opus 4.6 and 4.7 sit on trillions of parameters. Post-training is still compounding — RLHF, RLAIF, outcome-based fine-tuning are improving every release. Test-time compute is still compounding — reasoning modes burn more inference per query each generation. What changed is that we moved from one scaling law at a time to four scaling laws running in parallel, each feeding the next.

The new frontier is not which law wins. It is how tightly each law feeds the next. Whoever owns the loop between the layers owns the rate of capability gain.

Each Law Requires Its Own Substrate

The reason scaling laws stack rather than sequence is that each one feeds on a different kind of fuel — a different substrate to press against:

Pre-training fed on text. The open internet — a vast corpus of human-produced language that taught the model to compress the world into next-token predictions. When that fuel began to run out, the law did not stop compounding. It just became gated by something else.

Post-training fed on human feedback. RLHF, preference data, expert demonstrations — the second substrate was the alignment of model behavior to human judgment. Finite in a different way: bounded by the cost of generating quality feedback. RLAIF extended the runway, but the law remained substrate-gated.

Test-time scaling fed on verifiable reasoning. Chain-of-thought, problem decomposition, deep reasoning at inference. Math, code, science. The reason coding became the focal benchmark of the era is that code has the cleanest ground truth: it runs or it does not.

The fourth scaling law — the agentic loop — feeds on environments. Not text, not feedback, not reasoning problems. Environments. Actual computers, actual tools, actual outcomes the agent observes, retries, learns from. The fourth law cannot scale on benchmarks alone. It can only scale where the agent can act, fail, and observe consequences.

The Sandbox Is the Product

This is the architectural insight that ties the expansion to the scaling law. The sandbox the agent operates in is no longer a containment strategy. It is the training substrate.

The traditional reading of computer use is risk management — the lab “lets the agent out” of the cage just enough to do useful work, with guardrails to prevent misuse. That reading misses the structural point. The sandbox is not the cage. It is the environment in which the next generation of agents is being trained. Every Claude Computer Use session, every Perplexity Computer execution, every Operator browser run is a data point that feeds back into the next scaling-law iteration.

The cleanest formulation of this insight came out of analyzing Claude Cowork’s architecture: the sandbox is the new product. Not the output of the sandbox — the output is whatever the agent produced. The sandbox itself. Its design determines what the agent can learn, how fast it converges, and what next-generation capability it unlocks. The agent does not need freedom; it needs a wall to press against. Compute is not a cost — it is the substrate of convergence.

The wider the sandbox, the larger the surface area against which the fourth law can press. Sandbox expansion is therefore the binding constraint on the fourth scaling law.

Therefore Computer Use Is Not a Feature

Once the substrate logic is seen, the agentic expansion theory becomes inevitable rather than strategic:

Phase 1 (closed sandbox) was the substrate that pre-training fed on, externalized as a product surface.
Phase 2 (tool use) was the substrate that post-training and test-time scaling fed on.
Phase 3 (computer use) is the substrate the fourth scaling law requires.
Phase 4 (convergence) is the asymptote of the fourth scaling law — the limit where the agent’s training environment is indistinguishable from the world the agent operates in.

Every frontier lab is racing to ship Phase 3 and 4 not because the product market demands it but because the next capability jump depends on it. Without an expanded sandbox, the fourth law has nothing to climb on. Anthropic shipping Claude Computer Use on March 23. Perplexity declaring Personal Computer. OpenAI building agent-first hardware with Qualcomm. These are not competing product roadmaps. They are the same scaling-law substrate, instantiated differently.

This is the deep reason the agent absorbs the computer. The sandbox keeps widening because the scaling law requires it to. At the asymptote, the sandbox is the world and the agent is the computer.

Request Access To HQ Full Map Here!

The Subsidy Economy Is the Cost of Substrate Acquisition

The Map of AI Redrawn

Gennaro Cuofano — Sun, 24 May 2026 11:59:48 GMT

A decade ago, I started covering the intricacies of the AI ecosystem. As early as 2016–17, I described AI as a multi-layered stack, almost like a layered cake of interconnected technologies and infrastructure.

A few years later, right after the launch of ChatGPT, I began mapping the AI landscape in a far more granular way. The reason was simple: the ecosystem was finally crystallizing around a few critical building blocks.

More importantly, however, something I had emphasized for years became impossible to ignore. This entire ecosystem has been shaped by constraints, bottlenecks, chokepoints, and scarcity at every layer.

In an era where AI represents the next computing paradigm, this cascading ecosystem revealed how unprepared we actually were to make it work at scale.

From that realization, I developed the idea of the AI Supercycle: the notion that the AI industry resembles the semiconductor wave that began in the 1940s and 1950s, matured in the 1970s, and ultimately enabled the first great computing wave.

That wave gave us the Internet, the cloud, mobile computing, and all the sub-waves attached to it, from browser wars and search engines to social media and now AI super apps.

Interestingly, until just a couple of years ago, I believed it was sufficient to update the AI map once a year. Then reasoning models started to emerge, alongside additional scaling laws converging with pre-training. At that point, it became clear to me that the landscape needed quarterly updates instead.

Yet another shift is happening now. We are entering a phase where four scaling laws are simultaneously shaping the AI ecosystem, while several converging forces are radically transforming the physical layers underpinning AI itself.

And to be clear: while the physical infrastructure supporting the AI revolution will likely take more than a decade to fully mature, entirely new paradigms will continue to emerge on top of it in parallel.

For that reason, don’t be surprised if this map evolves rapidly. The AI paradigm caught most of us by surprise, and understanding it requires a high degree of mental flexibility as the landscape keeps shifting beneath us.

The May map described the geography — seven parallel competitive layers, each with its own clock. Seven weeks later, the architecture has revealed itself. The layers are still there, but the unit of analysis has changed. Companies stopped competing layer-by-layer and started binding multiple layers together through three distinct geometric strategies.

Three spines now cross the map. Two new buyer poles have emerged inside it. And one cascade is now flowing downward through four layers, changing the strategic logic at each level.

For Premium Members:

Get The Business Engineer Agent Here!

Try The Business Engineer Agent Here

The Nine Layers of AI

Gennaro Cuofano — Sat, 23 May 2026 11:59:50 GMT

In an era where AI represents the next computing paradigm, this cascading ecosystem revealed how unprepared we actually were to make it work at scale.

That wave gave us the Internet, the cloud, mobile computing, and all the sub-waves attached to it, from browser wars and search engines to social media and now AI super apps.

That is why the AI economy is not a list of companies racing each other. It is a layered industrial stack — and at any given moment, the binding constraint sits at a different layer than it did six months ago.

This research walks through all nine layers from bedrock to the surface, explaining what each layer is, why it matters right now, who controls it, and the mental model that governs its dynamics.

Read it as a field guide. Every headline, every valuation move, every strategic decision in the cycle lives inside one of these nine boxes.

For Premium Members:

Request Access To HQ Full Map Here!

The FDE Motion & The Unbundling of Consulting

Gennaro Cuofano — Fri, 22 May 2026 05:59:48 GMT

Management consulting has been the same shape for sixty years.

A team of bright generalists, organized into 8-20-week engagements, parachutes into a client. They run interviews, build models, produce a deck, present recommendations, hand off implementation to someone else, and leave.

The unit of work is the slide. The unit of value is the recommendation. The unit of billing is the partner-hour, leveraged through associates and analysts. The economics rest on a pyramid: junior staff doing the analytical work cheaply, billed at multiples that fund senior partner compensation in the millions.

This shape held when the work was understanding the business. When the constraint was what should we do? — Slides answered the question. The deck was the deliverable because the deck was the decision.

Enterprise AI broke that shape.

The constraint stopped being what we should do and became how we actually deploy this thing inside our workflows? The unit of work shifted from the slide to the running system. The unit of value shifted from the recommendation to the embedded artifact. The unit of billing — and this is the part the consulting industry is still digesting — shifted from the partner-hour to the engineer-week.

You cannot deliver a 5-day Bootcamp with a McKinsey associate. You cannot deliver a workflow-grade AI integration with a slide deck. The consulting motion optimized for the advisory layer is structurally misaligned with the deployment layer, and the deployment layer is where the value has moved.

This is the transition. Everything else in this piece is mechanism.

How SpaceX Is Rewriting the Map of AI

Gennaro Cuofano — Thu, 21 May 2026 11:59:47 GMT

Nvidia: The AI Map Maker

Gennaro Cuofano — Thu, 21 May 2026 06:29:48 GMT

NVIDIA — long described as the dominant chip vendor of the AI era — stopped being a chip company in the way that frame implies. It became the connector that holds the AI economy together.

The shift was visible in a quiet act of corporate editorial: the company rewrote how it segments its own business.

Two market platforms now replace four. Data Center —…

Google I/O 2026

Gennaro Cuofano — Tue, 19 May 2026 22:49:49 GMT

Google I/O 2026 was not a product showcase. It was a demonstration of coupling — every layer of the AI stack advancing simultaneously, each reinforcing the others. Read any single announcement in isolation, and it is incremental.

Read them together, and you see something structural: the full-stack flywheel completing its first revolution.

The Scale Context

Before the announcements, Sundar Pichai established the frame with three numbers.

Two years ago, Google processed 9.7 trillion tokens per month across its services. Last year at I/O that had grown to 480 trillion. Today: 3.2 quadrillion tokens per month — a 7× increase year over year. More than 8.5 million developers now build on Google’s models monthly. The APIs process around 19 billion tokens per minute. Over 375 customers each processed more than 1 trillion tokens in the past twelve months.

These are not vanity metrics. Tokens are the atomic unit of AI work — each one represents a fragment of a problem being solved, a question being answered, a task being automated. At 3.2 quadrillion per month, AI is no longer a feature being piloted. It is load-bearing infrastructure in production, at civilization scale.

The product numbers tell the same story. Google now has 13 products with over 1 billion users each — five of them above 3 billion. AI Overviews reached 2.5 billion monthly users. AI Mode, launched just one year ago, already has 1 billion monthly users with queries more than doubling every quarter. The Gemini app went from 400 million MAU at last year’s I/O to 900 million today, with daily requests growing seven times over the same period. Fifty billion images have been generated with Nano Banana alone.

The scale numbers matter for understanding the announcements that follow. They explain the $180–190 billion capex Google expects to deploy in 2026 — roughly six times the $31 billion it spent in 2022 — and they validate why investing in proprietary infrastructure at that level makes structural sense.

Interactive Map of AI Available Here!

Get The Business Engineer Agent Here!

FDEs & The Enterprise AI Adoption Wedge

Gennaro Cuofano — Tue, 19 May 2026 04:59:40 GMT

Enterprise AI has a problem that the model layer cannot solve.

The demo works. The pilot launches. The artifact never reaches the workflow it was meant to change.

Six months pass. The champion moves teams. Procurement freezes. The budget gets reallocated — and what should have been the deployment becomes another item in the graveyard of pilots that never …

AGaaS: The Five Architectural Primitives

Gennaro Cuofano — Mon, 18 May 2026 04:59:50 GMT

AGaaS — Agent-as-a-Service — is now the dominant category label for the AI-native enterprise software cohort of 2024–2026. Every analyst report uses it. Every founder pitch invokes it. Every fund deck cites it as the next architectural layer of enterprise software. The label has moved from contested terminology to default vocabulary in roughly eighteen …

The Business Engineer Agent

Gennaro Cuofano — Sun, 17 May 2026 16:39:51 GMT

There’s a question I’ve been asking myself for years: how do you scale a mind?

Join Directly Via The BE Platform

The Defensibility Map: From SaaS to AGaaS

Gennaro Cuofano — Sun, 17 May 2026 12:59:51 GMT

The February 2026 repricing forced a question every software operator and investor now has to answer: for any given software company today, what makes it defensible?

The instinct is to look for surviving moats. That instinct is wrong. There is no defensible SaaS in the old sense — only software that has migrated, or can migrate, into something else: an agent-shaped, outcome-priced, machine-callable execution layer.

Defensibility is no longer a position to defend. It is a migration to complete, measured against a capability frontier that is itself moving.

The Defensibility Map is the diagnostic for that migration. It is neither a SaaS framework nor an AGaaS framework. It is the framework for the vector between them — origin state, transitional state, destination state — and the system of layered diagnostics that determines whether a given company’s architectural position actually translates into commercial survival.

The Map produces five layered readings rather than a single score:

portfolio (capability-by-capability paradigm membership),
trajectory (where the center of mass is moving over 24 months),
adaptive capacity (positioning against the next frontier),
buyer-readiness gap (demand-side absorption), and
competitive geometry (relational defensibility within domain).

The diagnostic question shifts from “where is this company today?” to “where will this company be when the frontier moves, what’s the gradient connecting now to then, can the demand side absorb it, and who’s closest to eating their share?”

What follows is the architecture of that diagnostic — the structural physics underneath it, the six axes that score migration completeness, the three structural elements that turn a single-company score into a portfolio reading, the three composing frameworks that pair alongside it, and the methodology and rhetorical discipline that prevent the score from claiming more than it earns.

The Interactive AGaaS Defensibility Map

For Exec Members, The Full BE Agent Harness Is Now Available!

The Portal collapses seven assets I have built separately over the past decade into one substrate.

The AGaaS Race

Gennaro Cuofano — Sat, 16 May 2026 11:59:51 GMT

Back in March, I introduced you to the concept of AGaaS. Now it’s time to understand what makes it so.

Indeed, an Origin AGaaS is a company built natively, with no SaaS legacy to migrate from — it is the cohort the market is currently mispricing in both directions.

The bull case is that these companies are the structural winners of the AI paradigm shift. They started where every legacy SaaS company is now trying to arrive: outcome-priced, machine-callable, agent-operated, consumption-economic.

They compound with every additional agent deployed against them. Their growth curves are the steepest in software. They are the cohort the Defensibility Map’s “AGaaS-Native” position was named after.

The bear case is that most of them are wrapper plays whose entire product surface is a thin layer over a foundation model. Every Claude or GPT release narrows their wrapper.

Every hyperscaler agent platform announcement makes their distribution disadvantage more acute. Every incumbent figuring out the workflow brings the absorption threat closer.

By that reading, most Origin plays will be either acquired up into a larger platform within eighteen months or compressed into commodity API resellers.

Both readings are partially correct, and both readings collapse a critical distinction. Origin AGaaS is not a homogeneous winning cohort. It splits cleanly into two structural types, and only one of them survives the next platform shift.

Tooling-shaped Origin plays — companies whose product surface lives entirely above the foundation model — face a triple compression: foundation model capability eating them from below, hyperscaler platforms eating them horizontally, distribution-equipped incumbents eating them from above.

The eighteen-month window before one of those three forces resolves their position is the only meaningful clock they have.

Substrate-shaped Origin plays — companies that accumulate something durable underneath the agent layer (proprietary outcome data, regulated permissions, deep domain ontology, verification infrastructure, embedded workflow surfaces) — sit outside that triple compression.

The platform shift does not reach their moat, because the moat is not in the layer the platform shift commoditizes.

The Origin AGaaS Map is the diagnostic for which side of that line a given company sits on, and what determines whether they cross it.

The Interactive AGaaS Defensibility Map

The Portal collapses seven assets I have built separately over the past decade into one substrate.

Build Something Agents Want

Gennaro Cuofano — Wed, 13 May 2026 04:59:16 GMT

Paul Graham’s line — build something people want — was the operating instruction for two decades of product management. It assumed a specific topology: a human sees a product, evaluates it, decides, uses it. Every PM artifact — wireframes, user research, onboarding, retention cohorts — was built around that topology.

The topology is breaking. The evaluator, increasingly, is not a human. It is an agent: a model making a tool call, a planner selecting a sub-skill, an orchestrator routing a task. When the agent is the first-pass user, the PM stack inverts. Wireframes become schemas. Onboarding becomes discoverability in a tool manifest. Retention becomes reuse — does the agent call this tool again?

But the sharper version of the thesis — the one most takes miss — is this:

“Build what agents want” is incomplete. The real problem is designing systems where what agents want locally produces what humans want globally. That is an incentive architecture problem, not an AI problem.

The rest of this piece builds out why.

For Exec Members, The Full BE Agent Harness Is Now Available!

The Portal collapses seven assets I have built separately over the past decade into one substrate.

Interactive Map of AI Available Here!

For Exec Members, The Full BE Agent Harness Is Now Available!

Tokenomics: The Economics of AI

Gennaro Cuofano — Tue, 12 May 2026 11:59:49 GMT

Every conversation about AI in 2026 — capex, capacity, jobs, geopolitics, valuations, margins — eventually collapses to the same atomic unit. The token. It is simultaneously the unit of cognition the model produces, the unit of compute the data center serves, the unit of price the lab charges, and the unit of value the enterprise extracts. Almost no other commodity in modern economic history has played all four roles at once. Oil came close in the twentieth century. Tokens are doing it in the twenty-first, and faster.

This piece is a structural read of the resulting economy — the tokenomics — built up from the technical reality of what a token actually is, then cascaded back through the layered AI stack until every part of the ecosystem can be reanalyzed through that lens.

The Portal collapses seven assets I have built separately over the past decade into one substrate.

As part of it, as a founding member, you can also join the Alpha Program for the Business Engineer Agent.

Join The Alpha Program

Interactive Map of AI Available Here!

SpaceX: The AI Orbital Hyperscaler

Gennaro Cuofano — Mon, 11 May 2026 12:00:14 GMT

Three months in AI feels like ten years in the real world.

Just three months ago, Elon Musk publicly called Anthropic “evil,” “misanthropic,” and “hating Western civilization.”

Last week, Musk became Anthropic’s landlord — at a clip of $3-4 billion a year.

As you know, the Business Engineer doesn’t chase journalistic headlines. It looks for structural truths.

So the real question is: does this deal reveal any?

Quite a few, actually.

Follow along, because this might be more than a deal. It could redraw the entire CapEx Map.

On May 6, 2026, Anthropic signed an agreement with SpaceX to lease the entirety of Colossus 1 in Memphis, Tennessee:

220,000+ NVIDIA GPUs (H100, H200, GB200)
300+ megawatts of compute capacity
Online within a single month
Estimated $3-4B annual revenue for SpaceX, $2.5B+ in cash profit

The headlines have framed this as the personality reversal — the man who called Anthropic evil now monetizes its growth. That framing isn’t wrong. It’s just incomplete.

The structural story is that this single contract performs three different inversions at once. It changes Anthropic’s counterparty stack. It changes SpaceX’s revenue architecture. And it changes the cycle’s pattern of who-owns-what at the AI infrastructure layer.

By the time SpaceX’s S-1 becomes public — a confidential filing already made in April 2026, ahead of a June 2026 IPO targeting $1.75-2 trillion in valuation — this contract will have been re-read as something larger than its signing parties. It is, in operational terms, the first marquee customer of a fourth hyperscaler.

Whether that fourth hyperscaler thesis holds is the question the prospectus has to answer. And it’s the question that redraws the entire map of who participates in this capital cycle.

You can also find the interactive map below.

For Exec Members, The Full BE Agent Harness Is Now Available!

The Portal collapses seven assets I have built separately over the past decade into one substrate.

This is the same structural pattern that The Agent Manifesto identified at the level of individual practitioners: the consensus center is cheap; the right tail is expensive.

What the deal actually is

Strip away the personality reversal and the deal is mechanically simple.

The mechanics:

Capacity: All of Colossus 1 — Memphis, Tennessee, on the former Electrolux factory grounds in South Memphis
Hardware: 220,000+ NVIDIA accelerators (H100, H200, and newer GB200 systems)
Power: 300+ MW, supplied by Tennessee Valley Authority + Memphis Light, Gas & Water + on-site natural gas turbines (some of which operated without permits during 2024-2025)
Ramp: Full capacity available within one month — because the data center is already built and the GPUs already installed
Term: Undisclosed publicly. Reverse-engineered estimate from New Street Research: ~$3-4B annual revenue, ~$2.5B+ cash profit, with margin asymmetry coming from sunk capex (the $5B+ buildout cost is already on SpaceX’s balance sheet)

The two clauses outside the formal announcement that matter structurally:

One. In a reply on X, Musk wrote that SpaceX “reserves the right to reclaim the compute” if Anthropic’s AI “engages in actions that harm humanity.” Whether this clause appears in the actual contract is unconfirmed — but if it does, it gives Musk a soft control surface over one of the three frontier AI labs while he simultaneously runs a competing lab (xAI) and litigates against OpenAI in federal court.

Two. Both companies stated “interest” in jointly developing multi-gigawatt orbital data centers. This is not a current product. It is a stated direction — and the only direction in the cycle in which SpaceX has a structural advantage no other hyperscaler can match.

The trigger: a supply-demand mismatch large enough to bend ideology

At Anthropic’s developer conference in San Francisco the same week, CEO Dario Amodei said the company had planned for 10× growth in 2026.

Actual growth: 80× on an annualized basis in Q1.

That number deserves a pause. It changes the math of the cycle.

Anthropic’s annualized revenue: ~$3B in 2025 → ~$30B mid-2026 (10× year-over-year)
Within that, Q1 2026 alone delivered 80× the planned growth — driven by corporate adoption of Claude Code at Uber, Netflix, and Fortune 500 engineering teams building agentic workflows
The structural consequence: Anthropic’s existing compute commitments — already among the largest in the cycle — were calibrated to a growth path that turned out to be 8× too conservative

Anthropic’s existing infrastructure stack was already substantial:

The problem: all of these come online over years, not months. New gigawatts require land, permits, grid interconnects, and HBM4 supply (sold out at SK Hynix through H2 2027). What Anthropic needed in May 2026 was capacity now.

Across the cycle, every other operator was constrained. AWS, Azure, and GCP all reported AI capacity shortfalls in their Q1 2026 earnings. New construction lead times stretched 24-36 months.

The one operator with surplus capacity was the one nobody expected.

xAI’s Grok had been built with the same theoretical ambition as the leading frontier labs. Colossus 1 was constructed in 122 days starting June 2024 — an industrial speed record — to train Grok 3 and beyond. The expectation was that Grok would scale to a comparable revenue trajectory.

It didn’t.

The deal is the resolution of that mismatch. Musk has 220,000 GPUs and the highest-capacity AI training cluster on Earth. Anthropic has the demand he expected to have. The lease arbitrages the gap.

The structural patterns at work

The Rise of The Agent's Reward Designer

Gennaro Cuofano — Mon, 11 May 2026 04:59:10 GMT

The old PM was a user-cognition engineer; the new PM is an incentive architect.

Yet the reality is sharper and stranger.

The PM role isn’t migrating to a new center. It’s hollowing out in the middle and barbelling at the ends.

And it is the same pattern. The AI Orchestrator Playbook identified at the level of leverage: the model is a commodity; the scaffolding is the moat. Apply both lenses to the PM profession and a clear shape emerges.

The middle of the old PM craft — the thick belly of execution work, the journeyman tier where most PMs spent most of their hours — is exactly what swarms of agents are now eating.

Spec writing, ticket grooming, sprint coordination, status reporting, competitive teardowns, first drafts of designs, first drafts of analysis, first drafts of strategy. All of it can be generated, iterated, and executed by agent fleets faster than a PM can review the output.

The middle isn’t where the human PM has leverage anymore. It is where the human PM is the bottleneck.

What survives at the two ends is what agents cannot do, in two opposite directions.

At one end: taste. When agents can generate a thousand plausible candidates in an hour, the binding constraint stops being idea generation and becomes idea selection. The PM’s job at this end is to know which of the thousand is worth pursuing — and that judgment is the same instrument the Manifesto called direction. It is irreducibly human, irreducibly experiential, irreducibly slow to develop, and structurally beyond the agent’s reach because the agent’s prior is calibrated to Mediocristan, while the bets that matter live in Extremistan tails.

At the other end: sandbox design. When agents are doing the building, the testing, the iteration — when they’re running in the self-reinforcing loops the Manifesto called the agentic loop — what determines whether the loop converges on something valuable or spirals into reward-hacked nonsense is the quality of the environment you designed for them to run inside. The sandbox is not a metaphor. It is the operating condition the Playbook named sandbox + compute — the verifiable ground truth that lets the loop self-correct, and the iteration budget that lets convergence actually happen. The PM at this end is no longer designing the product. They are designing the conditions under which agents produce the product.

Both ends are versions of the same craft viewed from opposite sides. Taste is the orchestrator looking outward — selecting which problems deserve the loop in the first place. Sandbox design is the orchestrator looking inward — building the loop that will produce something worth shipping. Both are upstream of execution. Both are about distribution-reading. Neither is what the old PM craft trained anyone to do.

The PM who survives the agent era is the PM who is exceptional at one of these two ends. The PM who tries to hold the middle gets compressed out — not because they’re bad, but because the middle is where the leverage curve has gone vertical for everyone else and flat for them.

This piece is about both ends, why the middle is gone, and what the practical playbook actually looks like for each.

For Exec Members, The Full BE Agent Harness Is Now Available!

The Portal collapses seven assets I have built separately over the past decade into one substrate.

Interactive Map of AI Available Here!