We didn't gather in that Manila dormitory room in 2021 because we loved candlesticks. We gathered because a friend had lost three months of allowance to a contract none of us had read, and the silence in the room afterward felt heavier than the loss itself. That memory returned this week when a single line of market commentary crossed my desk: Goldman Sachs now forecasts that S&P 500 corporate profits will jump 27 percent, and that AI spending will carry the load.
Twenty-seven percent. The number is so large, so confident, that it reads less like a forecast and more like a confession. Because "carries the load" is not a phrase of abundance. It is a phrase of dependency. A structure that leans its entire weight on one beam is not strong because of the beam. It is fragile because of everything the beam is holding up alone.
We didn't spend the last decade building decentralized rails just to watch the most important economic claim of the year arrive without a single line of verifiable source code behind it. So let us do what we always do in this community. Let us open the ledger, examine the load, and ask the question nobody on the trading floor is asking: if AI is now carrying the load, who is carrying AI?
Context: The Forecast That Is Really a Structural Claim
To understand why this number matters, you have to understand what a 27 percent earnings growth forecast actually represents. The long-run average annual earnings growth for the S&P 500 sits somewhere between eight and ten percent. A 27 percent figure is not a good year. It is a structural claim. It says the machinery of American corporate profit has been rewired, and the new engine is artificial intelligence. That is a remarkable thing to assert in a sideways market, where price discovery has stalled, where capital is rotating rather than compounding, and where every serious allocator is quietly repositioning rather than celebrating.
In a consolidation phase like the one we are living through, headlines like this are not really about the next twelve months. They are about narrative. When price action goes quiet, the market does not stop searching for meaning. It simply transfers the search from charts to stories. And right now the dominant story in global finance is that artificial intelligence has become the load-bearing wall of the entire earnings complex.
There is something honest buried inside that phrase. If AI spending is carrying the load, then everything else is, by implication, not carrying very much. Consumer spending is soft. Traditional capital expenditure is cautious. Manufacturing is chopping sideways. The cyclical engines that historically drove broad-based earnings growth are idling, and one engine has been asked to pull the whole train. That is the real content of the Goldman headline, and it is a story about concentration, not about abundance.
For those of us who live in the crypto economy, this should feel familiar. We have spent years watching narrative concentrate value into a handful of assets while the long tail of the market bled out quietly. We know what it looks like when one beam holds the roof. We also know what happens when that beam develops a crack.
The difference is that in our world, the crack is visible. On-chain, a protocol losing forty percent of its liquidity providers in seven days is not a rumor. It is a number, timestamped, public, and irreversible. In the equity world, the crack in an earnings forecast is buried inside a model nobody outside the bank can inspect. That asymmetry is the reason I want to spend the rest of this piece doing something the headline will never do: tracing the physical and cryptographic layer beneath the AI economy, and asking whether the load it is carrying can actually be verified.
Because a forecast you cannot audit is not knowledge. It is a hope with a spreadsheet attached.
Core: The Layer Beneath the Machines
The AI economy is usually described in the language of software: models, parameters, tokens, weights. But the truth is that the AI economy is, at its foundation, a physical economy. It is electricity, silicon, copper, water, land, and heat. And above that physical layer sits a second layer that almost nobody talks about in earnings calls, but that everyone in this industry should be watching with total attention: the incentive layer. The layer that decides who gets paid, who gets verified, and who gets trusted when the actors doing the work are not humans but machines.
That is where blockchain stops being a curiosity and starts being infrastructure.
Let me be concrete about what I mean. Over the past two years, the center of gravity in my own work has shifted. In 2024, I spearheaded a pilot project that integrated a decentralized compute network with autonomous AI agents for content verification in the Philippines. We ran it with five developers and two sociologists, and we processed ten thousand data points, and we reduced misinformation in local news aggregation by roughly forty percent. That project taught me something the earnings models will never capture: the hardest problem in the AI economy is not compute. It is trust.
Because compute is becoming a commodity. Trust is not.
The single most important insight about the AI profit forecast is this: AI does not generate earnings by thinking. It generates earnings by being trusted to act. A model that drafts a contract is only valuable if someone can verify the contract was drafted correctly. An agent that moves money is only valuable if the movement can be audited. An algorithm that aggregates news is only valuable if the aggregation can be checked against reality. Every dollar of that forecasted 27 percent growth is ultimately a bet on verification.
And verification, at machine speed, is exactly what public ledgers were built to do.
Let me walk through the layers, because I think the market is systematically underpricing the middle of the stack.
The bottom layer is decentralized physical infrastructure. This is the world of distributed compute and storage networks, where idle GPU capacity is pooled, priced, and allocated through cryptographic coordination rather than through a single cloud provider's balance sheet. For years this category was treated as a hobbyist experiment. It is no longer. As AI training and inference demand accelerates, the marginal cost of compute becomes a first-order economic variable, and the ability to source that compute from a permissionless market becomes strategically interesting to anyone who does not want to be single-sourced.
Here is the part the equity analysts miss. A centralized cloud provider captures the AI capex story by owning the balance sheet. A decentralized compute network captures it by owning the incentive layer. These are different business models with different risk profiles, and the market currently prices them as if only the first one exists. That is a mispricing, and mispricings in a sideways market are where positioning actually happens.
The middle layer is the agent economy, and this is where I want to slow down, because it is the layer most likely to surprise everyone within the next eighteen months. As AI agents begin to transact autonomously, they need three things that traditional finance cannot easily give them. They need programmable money that settles in milliseconds. They need identity that is portable and verifiable across platforms. And they need payment rails that do not require a human to sign a form, open an account, or pass a compliance interview.
Traditional banking rails were designed for humans moving money at human speed. They break the moment you hand the keys to a machine. A machine does not fill out a wire transfer. A machine does not wait three business days for settlement. A machine does not have a credit history, a passport, or a phone number to verify. What a machine has is a private key and a set of instructions. That is the entire identity surface of an autonomous economic actor, and it maps perfectly onto the design assumptions of public blockchains.
We didn't build programmable money because we wanted faster gambling. We built it because we understood, earlier than most, that the next wave of economic actors would not be human. The 27 percent forecast is, in a very real sense, a forecast about machines becoming economic agents. And the moment machines become economic agents, the question of what money they use becomes unavoidable.
Now let me connect this to the part of the forecast that nobody wants to examine.
If AI spending is carrying the load, then AI spending is the most important variable in the entire market. Which means the market should be obsessing over a single question: what is the return on that spending? Here is where the forecast becomes fragile. A capital expenditure cycle and a revenue monetization cycle are not the same thing, and the gap between them is where every bubble in history has lived.
Consider how this actually plays out on the ground. A hyperscaler commits billions to data centers. The data centers require GPUs, cooling, power, land, and time. Those commitments show up as capital expenditure immediately and as revenue only later, if at all. For the earnings forecast to hold, the monetization has to arrive. And monetization in the AI economy depends on adoption, which depends on trust, which depends on verification. The chain of dependency runs from a bank's spreadsheet all the way down to whether a small business owner in Manila trusts an AI agent enough to let it handle her payroll.
That is not a hypothetical. That is exactly the terrain I have been working on since 2025, when I founded an academy to translate complex regulatory frameworks into accessible guides for small business owners, partnering with local banks on a curriculum for five hundred SME owners. The single most common question I received was not about returns. It was about safety. "Can I trust this?" "What happens if something goes wrong?" "Who do I call?" The AI economy will not scale on the strength of its models. It will scale on the strength of its guarantees.
And guarantees, at scale, require a trust layer. That is the layer the crypto economy has been quietly building for a decade, and it is the layer the equity market has not yet priced.
Let me make this concrete with the verification problem, because it is the intellectual heart of everything I am arguing.
An AI model can be confidently wrong. This is not a bug that will be patched away. It is a structural property of probabilistic systems. A model that is right ninety-five percent of the time is wrong five percent of the time, and in high-stakes contexts, five percent is catastrophic. In my 2024 pilot, we tested whether decentralized oracle networks could act as a verification layer to catch AI hallucinations in local news aggregation. The result was a forty percent reduction in misinformation, which sounds impressive until you realize it also means sixty percent remained. The point was never that the oracles solved the problem. The point was that they made the problem measurable, auditable, and attributable.
A hallucination that can be attributed to a specific model, a specific input, and a specific timestamp is not the same as a hallucination that vanishes into a private API. The first one is a manageable risk. The second one is a systemic one. And the entire 27 percent forecast rests on the assumption that AI systems will be trusted to act in high-stakes contexts. That trust cannot be manufactured by a bank's confidence. It has to be engineered.
This is where I want to bring in the sociological dimension, because it is the dimension that pure technical analysis always misses. Trust is not a cryptographic property. It is a social one. A ledger can prove that a transaction occurred. It cannot prove that the parties trust each other. What it can do is change the cost of betrayal. It can make defection expensive and honesty cheap. That is the real function of a public chain: not to create trust, but to make trust rational.
When we ran our DeFi resilience work during the 2022 winter, with two hundred members collectively auditing lending protocols and contributing findings to major projects, the thing that made it work was not the code. It was the fact that every contributor could see every other contributor's work. Transparency did not eliminate disagreement. It made disagreement productive. That is what a trust layer does at scale. It turns a room full of strangers into a system that can coordinate.
Now scale that up to an economy of autonomous agents. Millions of machines, transacting continuously, with no human in the loop, no legal recourse, no customer service line. How do you make that system trustworthy? You do it exactly the way we did it in that DAO. You make every action visible, every contribution attributable, and every betrayal costly. That is not a feature you bolt on after the fact. It is an architecture you build in from the beginning.
The AI economy is currently being built without that architecture. That is the largest hidden risk in the 27 percent forecast, and it is the reason I believe the crypto layer will become load-bearing in its own right.
Let me also address the physical constraints, because they are the part of the AI story that Wall Street consistently underestimates.
AI is a physical process, and physical processes obey physics. Data centers consume enormous amounts of electricity, generate enormous amounts of heat, and require enormous amounts of cooling. The marginal cost of inference is, at bottom, an energy cost. This means the AI profit forecast is quietly a forecast about energy, and energy markets do not respond to narratives the way equity markets do. You cannot talk a power grid into expanding faster. You cannot renegotiate the thermal limits of a transformer.
There is a reason the copper, power, and utility complex has become a proxy trade for AI exposure. It is because the AI economy is, at the physical layer, an energy economy. And here is where the crypto connection becomes surprisingly literal. The same capital that is being deployed into AI infrastructure is being deployed into energy infrastructure, and the same tokenization technologies that crypto has spent years developing are now being pointed at energy markets, compute markets, and real-world asset markets. The narrative is converging because the underlying physical reality is converging.
I want to be careful here, because this is where the omnichain-style marketing tends to take over. There is a strong temptation in our industry to claim that every macro trend is secretly a crypto trend. I have watched too many projects bolt the word "AI" onto their pitch deck and call it innovation. The test is never whether a project mentions AI. The test is whether the blockchain is doing load-bearing work that no other technology could do. If the answer is that a centralized database would work just as well, then the crypto is decoration. If the answer is that you need verifiable, permissionless, machine-readable settlement, then the crypto is infrastructure.
The distinction matters, and it matters more in a sideways market than in a bull one, because sideways markets punish decoration and reward substance. When liquidity is abundant, every narrative floats. When liquidity is scarce, only the load-bearing ones stay above water.
Which brings me to Bitcoin, and to something I have been saying for a while that the community does not always enjoy hearing.
When the spot ETFs were approved, a lot of people celebrated it as validation. I did not. I saw it as a transfer of custody. The Bitcoin that was supposed to be peer-to-peer electronic cash is now, in the dominant interpretation, a portfolio allocation line item held through regulated intermediaries on behalf of institutions. The instrument survived. The vision did not. Satoshi's original proposition, that two parties could transact without a trusted third party, has been quietly inverted. Now the most trusted third parties in finance hold the largest pools of the asset.
I am not being cynical. I am being descriptive. And the reason this matters for the AI forecast is that it reveals a pattern the market keeps repeating. Every time a decentralized technology reaches scale, the centralized world absorbs it, wraps it in a familiar instrument, and sells it back to the public with the decentralization stripped out. It happened to Bitcoin. It is happening to compute. It will happen to AI agents unless we are deliberate about preventing it.
So the real question raised by a 27 percent AI-driven earnings forecast is not whether AI will generate profits. It is who will capture them. If the answer is a handful of hyperscalers and their shareholders, then the forecast is a story about concentration. If the answer includes the open, verifiable, permissionless layer, then the forecast is a story about something more interesting: the emergence of a genuinely new economic substrate, one in which machines transact on rails that no single company owns.
That second story is the one I am positioning for. Not because I am certain it will happen, but because the asymmetry is enormous and the position is cheap. In a consolidation market, that is exactly the kind of bet you want to be making.
Contrarian: The Blind Spot in the Load-Bearing Narrative
Here is the counter-intuitive angle, and I want to state it plainly because I think almost everyone is getting it backwards.
The consensus interpretation of the Goldman forecast is bullish for AI. That is the obvious reading. AI spending is carrying the load, therefore AI is strong, therefore buy AI exposure. But there is a second reading that is far more uncomfortable and, I believe, far more accurate.
If one engine is carrying the entire load, then the forecast is not a statement of strength. It is a statement of fragility. The headline does not say that the economy is broadly healthy and AI is one contributor among many. It says that AI spending is carrying the load, which means the other engines have stopped pulling. Consumer spending, industrial capital expenditure, traditional corporate investment, all the diversified sources of earnings growth that historically made a broad index robust, are, by implication, weak. The index has become a concentrated bet dressed up as a diversified one.
This is the blind spot. When a market concentrates its earnings growth into a single driver, it also concentrates its downside risk into that same driver. The 27 percent number is impressive in the same way that a high-rise is impressive. It is impressive precisely because it is tall and thin and standing on a single foundation. Nobody celebrates a tall building for being structurally diversified. We celebrate it for being tall, and then we worry, quietly, about the wind.
And the wind is coming from a direction the model cannot see. The entire forecast depends on AI capital expenditure continuing at its current pace. But capital expenditure is a decision, not a law of nature. The moment a few large buyers decide that the returns have not materialized, the spending slows, and the load-bearing beam develops a crack. The forecast has no answer for this, because the forecast is the beam.
There is a deeper issue here, and it is the one I keep returning to. The 27 percent figure is a prediction, not an observation. It rests on a model whose assumptions we cannot inspect. In our world, when a protocol claims a 27 percent yield, we ask to see the contracts. We ask who audits them. We ask what happens in the tail. The equity market asks none of these questions, because the model is proprietary and the confidence is institutional. A forecast you cannot audit is not a forecast. It is a positioning statement wearing the costume of analysis.
So my contrarian conclusion is this: the Goldman headline is not a signal to buy AI exposure. It is a signal to understand what happens when the load shifts. The real opportunity is not in the beam that is currently carrying the weight. It is in the foundations being poured underneath it, quietly, by people who understand that the next economic substrate will not be owned by any single company. That is where I am looking. Not at the headline. At the layer beneath it.
Takeaway: What We Are Actually Building For
We didn't come this far to build a machine economy that simply reproduces the concentration of the old one, with a few hyperscalers replacing a few banks and calling it progress. The forecast tells us that machines are about to carry the economic load. It does not tell us who will own the rails they run on. That question is still open, and open questions are where builders do their most important work. The load is shifting. The question is whether the trust layer shifts with it, or whether we let the weight settle, once again, into the hands of the few. Everything we build from here is an answer to that.