Blackstone President and COO Jon Gray: We Are Closer to the Eve of the 1870 Industrial Revolution, Major Industrial Revolution-like Changes Are Coming
Jon Gray, President and COO of Blackstone, stated in a presentation to over a hundred investors that the current situation is closer to 1870 than the 2000 bubble: major industrial revolution-like changes are coming. Blackstone subsequently released the full speech titled "Where's the Beef?".
He cited that from 1870 to 1900, U.S. labor productivity doubled, GDP increased about fourfold, real manufacturing grew about sixfold, and the stock market rose about sevenfold, stating that this is not a prediction of today's growth but rather an illustration of the scale of technological diffusion. On the model company side, he mentioned that Anthropic and OpenAI had almost no revenue a few years ago, but by a certain July, revenue expanded about 100 times, showcasing an annualized revenue of approximately $105 billion, and noted that such speed is rare.
He used the portfolio ledger as validation. Among about 1,400 invested companies, holdings, and borrowers, Anthropic's annualized expenditure increased from about $25 million last September to about $525 million, approximately 21 times. He said that nine out of the ten highest-grossing investments for the company that quarter were concentrated in artificial intelligence. The capital expenditures of five major cloud companies doubled from about $415 billion last year to about $820 billion, equivalent to about 2.5% of U.S. GDP.
Blackstone's own data center platform is expected to lease about 1 GW in 2024 and about 2 GW in 2025, with a target of at least about 6 GW that year, corresponding to nearly $100 billion in capital expenditures for the data centers, and tenants are expected to invest about $200 billion more in chips. He compared SK Hynix's approximately 4 times price-to-earnings ratio with Cisco's approximately 150 times in 2000, stating that today energy and data centers have demand exceeding supply, with contract counterparties having low leverage, unlike the over-leveraged railroads of the past.
Micro examples include a communication tower company investing about $4 million in AI processes, generating annual revenue of about $4.5 million, and the energy data company Enverus demonstrating a project with about 18 times return. U.S. labor productivity has risen from an average of about 1.5% over the past decade to nearly 2.6% over the last two and a half years. Physical bottlenecks were specifically pointed out: General Electric's Vernova turbines are queued until 2031, costing about $55 billion per GW of power. Retail comparisons include Kmart and Sears exiting, while Walmart and Costco survive. The strategy is to bet on foundational chips, data centers, and power, as well as results-oriented record systems.
The market mechanism is the allocation of infrastructure rights. Buyers are model companies and cloud providers looking to lock in machines, while sellers are alternative asset managers holding land, power queues, and leasing platforms. Event-driven opportunities arise from official releases. Beneficiaries are data centers and credit packaging with long-term leases, while those under pressure are equity themes that still view AI as a purely software story without physical positioning. Funds are flowing from model subscription fees to cement, turbines, and custom chip leasing.
Source: Public Information
ABAB AI Insight
Gray is not a passive commentator. Blackstone privatized QTS in 2021 to enter the data center market, then scaled leasing to the gigawatt level, and along with Apollo and Broadcom, expanded Anthropic's capacity by about $35 billion. The "proprietary data" mentioned in the speech essentially refers to the bills of tenants and borrowers: the willingness of portfolio companies to increase spending from $25 million to $525 million indicates that costs have shifted from experimental budgets to operational expenses. He shifted the reference point to 1870 to exclude the 2000 price-to-earnings ratio from the discussion.
The capital path is clear. Equity buys the growth of model labs, credit buys the depreciation of chips and cabinets, and real estate buys the power queue. Blackstone claims it wants to be the world's largest investor in AI infrastructure, with internal data center valuations mentioned to be around $185 billion by Q2 2026. Money is not flowing to "who's the next chat window," but to those who can secure turbines and substations before 2031.
The analogy is not to the Cisco bubble, but to the railroad era's ties and telegraph companies: some lines were overbuilt, while others became standard gauge. Today's equivalents are megawatt leases and custom ASICs. The industry is in a phase of expansion and positioning overlap: the demand curve remains steep, and supply is locked by the power grid, thus rents and electricity prices are being priced before software profits.
Structural changes are reconstructing the industrial chain. Intelligence is being made into subscriptions, while the physical layer is being turned into long-term rentable assets. The mechanism is: model revenue proves someone is paying, capital expenditure proves money has turned into irreversible concrete; whoever controls power and cabinets will charge model companies tolls. When the electric light of 1870 was not yet widespread, copper wires and power plants were already beginning to collect rent.
ABAB News · Law of Cognition
Bubbles fear oversupply, industrial revolutions fear power queues
Software tells stories, infrastructure tells leases
The real beef often lies in the least sexy process transformations