We Are at the Dawn of 1870 [Full Transcript of the Blackstone CEO’s Closed-Door Speech]

Companies like Anthropic and OpenAI, which had almost no revenue just a few years ago, have seen revenue grow 100 times by July 2025. As of July, their annualized revenue has reached $105 billion. This is unprecedented in human history.

These two sentences come from Jon Gray, President and COO of Blackstone Group. On September 20, during a closed-door presentation to over 100 top global investors, he disclosed for the first time Blackstone’s proprietary data and core judgments on AI investments. Below is the full transcript of his speech.

Part 1: Opening

I really like this music. This is Blackstone’s 2024 holiday video, and the 2026 video is coming soon. I’m delighted to have you all here from around the world—thank you very much. Without your support, we wouldn’t be where we are today. We’re always grateful. Thank you.

Part 2: Starting with the Most Important Question

Now, we’d like to address the key concerns you all have when considering making massive investments. I think it’s best to start with the most important question: How to become a thought leader in the global financial fitness space.

This is very important to me. I have five tips here.

First, you need to sweat profusely. Fortunately, I’m a natural sweater, but it makes you look very authentic.

Second, you need to angle the camera upward. My daughters always remind me to do this—otherwise, you get a terrible double chin.

Third, you should show your brand without hesitation. If you want to feel like a NASCAR driver, display a logo that says you’re important.

Fourth, you need a “goofy dad” vibe. Perfect. Add five emojis—that shows full enthusiasm and makes filming fun.

That’s enough buildup. Let’s get to the main topic.

Part 3: The Four Questions We’ll Address This Morning

This morning, we’ll try to answer four questions.

First, what’s changed since most of us were last together last year?

Second, what can we learn from the past, especially the distant past?

Third, where are we today?

Fourth, where are we heading, and what will happen in the future?

Part 4: The AI Perspective

We’ll, of course, approach this from the two letters of AI. This doesn’t mean we’re not concerned about geopolitical tensions in Ukraine and the Middle East—these have pushed up commodity prices, driven inflation, and raised short-term rates, as we saw yesterday. People are worried about deficits, long-term rates—all important for investors. But when we think about what ultimately drives portfolio value, what’s happening in the economy, and what will happen in the future, we must consider these two letters.

So, let’s go back to one year ago—September 17, 2025—when we were in this room. What were we thinking then?

What was our primary reason for staying enthusiastic? Mainly, it was AI. Our style hasn’t changed much in the past year.

Part 5: Reflecting on the Past Year

When we look back at the past year or two of AI development, take Google tokens as an example: trading volume went from nearly zero in 2024 to 480 trillion tokens per month today. By May 2026, it had reached 3.2 quadrillion tokens—a new term I encountered while preparing this speech. By September, these numbers will clearly be even higher.

Companies like Anthropic and OpenAI, which had almost no revenue just a few years ago, have seen revenue grow 100 times by July 2025. As of July, their annualized revenue has reached $105 billion. This is unprecedented in human history.

Blackstone has a unique perspective through our portfolio companies. These data come from our portfolio companies, GPs’ portfolio companies, and our borrowers. Last September, these 14 companies had annualized revenue of $25 million. Now, it’s grown 21 times to $525 million. This reflects the current market, as these companies are earning exceptional investment returns.

This obviously affects valuations. We’re fortunate to have invested in both of these companies. Their combined valuation last September was $683 billion, and now it’s estimated to exceed $2.3 trillion. As they begin to go public, their strong revenue models are driving valuation growth.

Part 6: Infrastructure, CapEx, and Power

We also discussed a lot last year: not just models and technology, but the infrastructure needed to support them. Global CapEx has grown six times, and these numbers will be even larger next year. Last year, the five hyperscalers spent $41.5 billion on CapEx. This year, that figure has doubled to $82 billion, equivalent to 2.5% of U.S. GDP.

It’s not just hyperscalers—we’re actively involved in data center leasing. We have the world’s largest platform. We leased 1 gigawatt in 2024, 2 gigawatts in 2025, and expect to lease at least 6 gigawatts this year, representing nearly $100 billion in CapEx, with tenants also investing hundreds of billions in chips.

Unfortunately, we didn’t invest in memory chip companies. Leading players like Micron and SK Hynix saw stock prices surge 600% and 500%, respectively. Looking at such charts, investors’ first reaction is often: this looks like a bubble. But SK Hynix’s current P/E ratio is only 4 times—not like Cisco in 2000, when it was 150 times. The market is still quite skeptical about the current situation.

Last year, we invested heavily in GPU and TPU financing, partnered with Nvidia and Broadcom, made big investments in new cloud computing areas, and formed a joint venture with Google. Through Crux, we invested with Firmus in Australia, NASA and India, and many data center companies worldwide. We believe strong demand for computing power will continue to yield rich returns.

It’s not just data centers—power is increasingly important. Last year, we talked about the critical role of power for the future. Over the past five years, utility CapEx has totaled $800 trillion, and it’s expected to nearly double in the next five years. I’m visiting Canada this week, where they’re discussing investing hundreds of billions to expand and upgrade the power grid; similar situations are emerging globally.

Stocks in this sector haven’t reacted as dramatically as memory chip companies, but the power industry was once considered dull. Now, companies making turbines, cooling equipment, and electrical gear have seen stock prices jump 44% to 81% in a year. Fortunately, we’ve also invested heavily here—we provided a huge loan to Sempra Infrastructure to fund a large LNG project; invested in energy data analytics company Enverus; invested in Williams to build utilities, power plants, and data centers, including the McLane project; and invested in European electrical equipment, wind and solar technology companies, and renewable energy firms. We’ve made massive investments in energy transition and infrastructure businesses.

What returns have these AI-related investments brought? We analyzed the ten highest-yielding investments company-wide for Q2 this year, and nine were in AI, including some industry benchmarks—only India’s Aster Care hospital was an exception. We expect Q3 to be very similar.

Part 7: Learning from the Distant Past: 1870–1900

So, what can we learn from the past, especially the distant past? Let’s go back to September 17, 1870. Most Americans lived on farms then. It was post-Civil War America, and almost all buildings were made of wood. To visit relatives in the next town, you’d ride a horse; to read at night, you’d light a candle.

In just 30 years, everything changed. By 1900, many buildings were made of steel; for visits, you could take a train, and cars began appearing on roads; thanks to Thomas Edison, people could read under electric light.

I think our situation today is closer to 1870—a major industrial revolution is about to unfold. From 1870 to 1900, what happened economically? Annual labor productivity doubled, GDP grew four times, real manufacturing output increased six times, and the stock market rose seven times. I’m not saying the same thing will happen today, but it does show the power of technological change and innovation—and what they ultimately mean for the economy and investments.

But even during booms, recessions happen. The key infrastructure then was railroads, and during that period, 200 railroad companies went bankrupt or insolvent. This is a warning. Back then, they built supply ahead of demand with high leverage—much like the later telecom infrastructure build-out. Today, it’s different: demand for energy and data centers far exceeds supply, and most commitments are from companies with very low leverage. This doesn’t mean there won’t be problems, but the situation has changed.

Part 8: The 1984 Ad and “Where’s the Beef?”

Now, let’s fast-forward 84 years. On January 22, 1984—just before my 14th birthday—I was at Elm Place Middle School in Highland Park, Illinois, watching the Washington Redskins, coached by Joe Theismann, take on Marcus Allen’s Oakland Raiders. The game wasn’t exciting—38–9. But two ads stood out.

The first was Steve Jobs’ famous 1984 ad: “On January 24th, Apple Computer will introduce Macintosh. And you’ll see why 1984 won’t be like 1984.”

This ad was prophetic—it was about making computers accessible. People shouldn’t be forced to use Cobalt or MS-DOS with all those complicated commands; they should just click a mouse and have powerful computing at their fingertips.

If we compare it to today, we’re talking about everyone in the world having access to intelligence—even in the smallest villages, people could tap into the world’s greatest minds without being in Cambridge, UK, or Cambridge, Massachusetts. That’s a powerful message.

But I think the most relevant topic from this discussion is the question Clara Peller asked during the same Super Bowl. This 81-year-old woman looked at a giant fluffy bun and asked a very important question: “Where’s the beef?”

That’s exactly the question we’re facing: We’re investing trillions in CapEx—what does it really mean? Does it make sense, or is it just circular financing that will ultimately cost people everything?

Part 9: Current Status and AI Applications

This is what we’re trying to answer. I want to return to the current situation and explain why I’m confident in these investments. It all starts with what’s happening inside our companies: AI use cases are expanding rapidly—almost everywhere.

For example, in process improvement, there’s India’s IGI diamond grading system and our mobile tower company Phoenix Tower, which is now processing leases five times faster. The company invested $4 million in new AI processes and now gains $4.5 million in annual benefits—a software engineering ROI of over 100%. Almost all our partner companies are benefiting. Enverus has built a new code remediation process, and its investment in the demo company has yielded an 18x return—that’s why investment keeps growing.

In customer service, Tricon, a rental housing company, can speed up application processing by 90%. The same goes for content creation, advertising, design, and rendering.

But this isn’t just about productivity. Our portfolio company Chamberlain, which makes garage door openers, launched the MyQ Secure View 3-in-1 smart lock with facial recognition. A few years ago, they realized they could use AI visualization to expand into front doors and Amazon package delivery via garage doors. Now, their annualized revenue is $40.5 million, and the CEO expects it to reach $500 million annually within five years. Over time, we’ll see more and more such cases.

In legal and compliance, marketing review efficiency has improved by 50%. In software development, what I personally appreciate most is the CIO and portfolio intelligent agents, which reduce time by 99%.

Now, these advantages are showing up in data. U.S. annual productivity growth averaged 1.5% over the past decade but has reached 2.6% over the past two and a half years. Hyperscalers are leading this trend—their revenue per employee has grown 65% over the past three and a half years, and profit margins are rising. EBITDA margins for S&P 500 companies and our portfolio companies have improved by 500 to 700 basis points over the past four years.

This has driven strong earnings growth. The average growth rate over the past 12 months is 15%, and even excluding one-time gains, it’s 32%. There’s no doubt that as this technology spreads through the economy, it will have a massive impact.

Part 10: Technology Reaches the Real World

Now, everything will change. Currently, most of these technologies are still on desks. But as they move off desks and into robots, autonomous vehicles, national security, and other real-world applications, demand for computing power will keep growing, and productivity will continue to rise.

Take autonomous vehicle company Waymo: Over the past two and a half years, its mileage has grown 250 times, while severe accident rates have dropped by 94%. This trend will continue and have profound effects.

Regarding AI’s impact on our lives, challenges are inevitable, and society will undergo transformation. But it’s also meaningful to discuss some truly positive effects.

One is job growth. At our data center company QTS, blue-collar jobs are booming—in less than two years, the number of construction workers on-site has tripled. Despite AI boosting productivity, our company’s overall employment is still growing. Over the past decade, the number of new business applications has doubled, and starting a company is much easier now.

Part 11: Science and Health

The most exciting part of this technology is science. A study published in Nature showed that drugs discovered through AI have significantly higher success rates in Phase I clinical trials. AI is definitely superior in chemistry; in biology, it still needs improvement, but the future looks promising. A recent McKinsey study found that AI optimization can shorten trial duration by 40%. Our portfolio company Vara is doing the same in this field. For everyone, the most exciting thing is: How will AI improve our health long-term?

Part 12: Three Core Beliefs for the Future

Given all this is happening, where should we go? We have three core beliefs.

First, these use cases will proliferate, boosting productivity dramatically—not just in healthcare and legal sectors, but in the real world. As productivity rises, people will use AI more, driving exponential growth in demand for intelligence. We believe this will continue. The real bottlenecks are in the real world—in chips, power, and data centers.

Part 13: Real-World Challenges

Why do we face these challenges? I’m standing on a data center construction site—a literal AI factory—and there are tangible hurdles here.

First is equity concerns—communities are worried. Unfortunately, there are many misconceptions: building data centers uses minimal water, benefits communities, and uses new energy to lower power costs. But concerns will still slow projects—for example, we’ve seen project pauses in New York State.

Second, power construction is difficult—getting a turbine from GE Vernota might mean waiting until 2031.

Third, in chip supply, hyperscalers have increased their capacity cap nine times in five years, while chip companies haven’t even doubled theirs. The chip industry itself is cyclical, and companies have been hesitant, leading to severe shortages.

There’s also an issue crucial to us: These AI factories, including power, data centers, and chips, require huge capital—$55 billion per gigawatt. Blackstone has the unique scale to understand the situation and mobilize needed resources.

Now, looking back at last year’s biggest worry: Underestimating the risk of technological disruptive innovation.

Last year, this showed up in the stock market—professional services, software, and information services companies saw stock prices drop. In many cases, their core business was healthy, but market participants worried about future prospects, uncertainty increased, and valuation multiples compressed. This also happened in private markets—software deals fell by 66%, with larger deals dropping even more.

But not all companies face the same outcome. The six retailers that confronted Amazon’s impact 25 years ago are a good example: Kmart, Sears, and Toys R Us eventually faded, while Walmart, Costco, and TJ Maxx survived. They had unique value propositions and weathered the storm.

Similar dynamics will occur in white-collar firms. Some companies have critical record systems, management teams shifting focus from seats to outcomes, and ultimately delivering better performance. During this time, some companies may have lower valuation multiples due to uncertainty.

Part 14: Scarcity Assets

There are also winners unrelated to tech—let’s call them scarcity assets. We invested in an Indian cricket team. India has 1.4 billion people, and cricket is the most popular sport. Currently, there are only 10 teams, and AI can’t replace cricket. We also invested in a great coffee chain, 7 Brew, whose signature drink is the Blonde Roast. People still need and love these products.

Waterfront property might be one of the best investment areas. Despite real estate being out of favor with low valuation multiples, these experiences won’t disappear—the assets are irreplaceable. Or infrastructure like Rome Airport, which is also critical.

Part 15: The Biggest Risks

So, what are the biggest risks?

First, cybersecurity. A failure in critical financial infrastructure could trigger political turmoil, not just in the U.S. but globally.

Second, if regulation slows the tech frontier, it won’t change the inevitability of technology diffusion—because today’s tech is already very powerful. Continuous technological advancement itself is a significant risk. Things like space data centers and edge computing on phones could happen. But given the demand curve, this is somewhat like power: We need nuclear, natural gas, and renewables to develop in parallel. Currently, ground-based data centers are still very important.

Third, geopolitical tensions. Fourth, overvaluation. Companies with little to no revenue are valued at $10 billion, and some defense tech companies have astonishing valuations—all need close monitoring.

These are the issues in our daily investment discussions, and why we focus on the bottom of the tech stack—computing—because we’re most confident in this part.

Part 16: The Core and Returns

So, what’s the real “core”? It’s chips, data centers, and power. And above that, the real core is returns on investment. We’ll continue to see more use cases, sustained demand growth—and that’s what justifies this massive investment.

Finally, we’ll end with that “North Star” slide—as we often say, “It’s all about returns.” Since Pete and Steve founded this company 40 years ago, we’ve been dedicated to creating value for our clients. Steve reminds us of this every day. That’s why we write memos, read them on weekends, and spend so much time thinking about the world’s direction.

Gratitude is an enduring value. We’re grateful to all of you here. I hope you enjoy the rest of your day. Thank you again.