When Governments Become AI's Venture Capitalists

History doesn't repeat, but capital cycles often rhyme. AI may be entering its next chapter.

KEY TAKEAWAYS

• Every major technology innovation follows a recognizable capital cycle that begins with visionary venture capital, progresses through institutional and corporate investors, and eventually attracts larger pools of capital seeking established opportunities. Tracking who is funding an industry often reveals where it sits in its maturity cycle.

• The creation and rapid growth of MGX illustrates that sovereign wealth funds have become major providers of capital to the AI ecosystem. Their participation suggests AI infrastructure financing has progressed beyond traditional venture and growth capital.

• AI infrastructure spending continues to accelerate even as revenue growth remains far lower than capital expenditure growth. Rising debt issuance and alternative financing structures indicate that internal cash generation is no longer sufficient to fund expansion.

• Unlike many traditional infrastructure assets, AI hardware requires continual replacement as technology advances. A growing share of AI investment now represents maintenance spending needed simply to remain competitive rather than expanding capacity.

• Historical infrastructure booms demonstrate that valuable assets can survive while many of the companies financing their construction do not. Investors should distinguish between the long-term success of AI infrastructure and the investment outcomes of individual participants.

MY HOT TAKES

• Capital cycles provide a more reliable framework for identifying investment opportunities than media attention or popular narratives. Following the source of funding often reveals more than following the technology itself.

• Sovereign wealth participation should be viewed as evidence that AI financing has entered a more mature stage rather than as proof of limitless demand. The character of capital supporting the industry has fundamentally changed.

• Investors should pay closer attention to AI companies' balance sheets and cash flow than to headline spending announcements. Large capital expenditures eventually require sustainable economic returns.

• The next generation of outsized investment winners is likely to come from businesses monetizing AI infrastructure rather than simply participating in its construction. Profitability ultimately matters more than technological novelty.

• The AI boom may continue for many years, but exceptional returns become increasingly difficult as more capital floods into a well-established investment theme. The greatest opportunities often exist before widespread public enthusiasm develops.

• If you really want to know what’s up, ask my wife–she knows. 😉

• You can quote me: "AI isn't running out of innovation just yes–but it is running into the realities of capital."

The old, new thing. I have been at this for nearly four decades. 😅 Through it all, I have been lucky to have a partner–my wife–by my side. Let’s establish something up front. She KNOWS what’s up–she has a talent for recognizing wind changes in business. She is also indefatigable (one of my favorite words–also the name of a legendary British ship of the line ⚓). If she senses something she presses until she gets the point across. Got it? Good. Well just the other day, she was pressing me on “what’s the new, new thing,” spurring me to think critically–essentially tear up the comfortable narrative–or at least check it and make sure it is still relevant. She asked the question, but she was already prepared with a multiple choice answer. My work was cut out for me. I went down her list explaining that while many of those choices were certainly new and exciting, there was a certain nuance–oft misunderstood–that distinguished an idea or technology that was just “new,” from one that was “new, new.” She persisted and pushed back, urging me to give proof of my dismissals. I explained that a good litmus test for the state of “newness” was to follow the money. More specifically, who is funding it.

I will refrain from describing this as smart versus dumb money, but now that it's out there…well, it’s out there. You see there are a group of folks that invest in far-out ideas. Amongst them, some are really skilled, while most are just rainbow-chasers. You want to watch that small group of skilled, early funders. That, of course, is not always easy–but there is still plenty of opportunity, once the idea/tech moves to the next level. These folks are also very smart but they have a lower bar for acceptance–they rely on the smart far-out investors to narrow the field and cull out the obvious non-starters. Here too, there are skilled and less-skilled. Obviously the skilled are the ones to watch. After this group comes the heavy lifters. They have money and they are good at sniffing out short-term win opportunities. This is the group that is easier to watch, but be careful, only a small minority of them are actually skilled and have a knack for success. After this group gets involved there is only dumb, opportunistic money whose only skill is to fill short-duration cash flow needs hoping to bridge the idea to a successful exit, which is–these days at least–a star-studded IPO. Interestingly, once the first IPO happens and hits the mainstream news, a fresh cycle of copycats quickly emerges to follow in the wake of the first. Those followers have very little chance of achieving the returns of the first movers, but fret not, there are lots and lots…and lots of funding entities looking to pour hot money into what they perceive as being the new, new thing. The far-out investors from above 🙃–the ones at the beginning of the chain–well, they have already moved on.

My friends, I can assure you that once you hear about the next big thing at the family get-together or on your favorite mainstream financial media channel, it is no longer “new,” nor is it “big.” That doesn’t mean you can’t make money investing in it, but the great returns window has already slammed shut.

So, I am sure you are wondering, where does AI fit in that framework? Let me tell you about a two-year-old firm you probably have not heard of–until this week.

MGX is a sovereign investment vehicle born in Abu Dhabi in March 2024. It was created by the Abu Dhabi government, backed by Mubadala, which is one of the world's most sophisticated sovereign wealth funds, and G42, a UAE-based AI holding company. Its chairman is Sheikh Tahnoon bin Zayed Al Nahyan, the UAE's National Security Adviser and a brother of the country's president. In under thirty months, MGX has co-led Anthropic's $30 billion funding round, co-led OpenAI's $122 billion raise, joined the Stargate consortium, backed Elon Musk's xAI, and participated in the $40 billion acquisition of Aligned Data Centers–one of the largest private equity infrastructure deals ever recorded. This week, MGX announced it has closed its Fund I at $49 billion, incidentally exceeding its initial $45 billion target, drawing capital from regional sovereign wealth funds alongside global pension funds and institutional investors from North America, Asia, and Europe. The firm is now targeting over $100 billion in total assets under management.

The tech press called it validation. A standing ovation for the AI boom. Proof that the money never stops.

I called my wife.

The five largest U.S. hyperscalers–Microsoft, Amazon, Alphabet, Meta, and Oracle–have collectively committed somewhere between $660 and $690 billion in capital expenditure for 2026 alone, which is old news by now. To put that in perspective: in 2024, the combined capex of just the four biggest was a little over $200 billion. In two years, it has more than tripled. The AI arms race now represents what some analysts are calling the single largest corporate capital expenditure cycle in recorded history.

And yet, CAPEX is growing at roughly 80% while revenues at these same companies are growing at 15 to 16%. Free cash flow is deteriorating. These firms, which largely ignored the debt markets for a decade, issued more new debt in the first quarter of 2026 alone than they did in all of 2025. You also know, if you read yesterday’s blogpost/newsletter that a whole lot more was raised by this group through off-balance sheet transactions (https://blog.siebert.com/ais-biggest-risk-isnt-technology-its-credit). The internal cash generation machine has been outrun by the infrastructure ambition. The balance sheets are showing it quietly, in the footnotes, the way these things always show up first.

There is another wrinkle that rarely gets discussed. AI hardware does not age gracefully. It does not sit in a data center for twenty years the way a turbine sits in a power plant. The GPU generations are turning over so rapidly that a meaningful portion of hyperscaler AI capex is not growth CAPEX–it is maintenance CAPEX. They are running on a treadmill, replacing equipment just to hold their competitive position. One research firm put it bluntly: these companies are less like traditional technology enterprises and more like supermarkets–constantly restocking shelves, just with billion-dollar chips instead of groceries.

So follow my framework with me. The early far-out money–the visionary venture capitalists who seeded OpenAI, Anthropic, and the foundation model layer–they made their bets years ago. Many of them have already moved on, exactly as I described to my wife. The second tier–the heavy lifters, the sophisticated growth funds, the strategic corporate investors–they poured in next, pricing their entry on the expectation that the early bets were validated. Now we have arrived at the final stage of private market funding. And here is where MGX becomes the most important data point of the week.

When a sovereign government has to stand up a purpose-built, $49 billion investment vehicle just to keep the capital flowing into a two-year-old technology cycle, that is not a sign of endless abundance. That is a sign that private market liquidity may have hit its ceiling. The funding baton has been passed from venture capital to growth equity to corporate balance sheets–and now, finally, to nationalized sovereign balance sheets that answer to governments rather than shareholders. The AI infrastructure build is now, in a very real sense, being backstopped by states.

That is not necessarily fatal to the thesis. Sovereign capital is patient. It does not face the same quarterly redemption pressures that a traditional fund manager does. But it does change the risk calculus for retail investors watching from the sidelines. Because when sovereign governments become the underwriters of last resort for a private technology sector, the exit dynamics change. The IPO cycle that follows–and it is coming–will look very different from the venture-backed, organically-grown public offerings of prior technology generations.

I have been around long enough to remember when the telecom buildout of the late 1990s attracted its own wave of increasingly desperate, hot capital in its final innings. The infrastructure they built was real. The fiber in the ground was real. The carriers that overbuilt it–the ones who needed that last tranche of bridge money–most of them did not survive to see the infrastructure they built become profitable. The infrastructure outlasted the equity. I am not saying AI ends the same way. The use cases are more immediate, the monetization more tangible. But the capital cycle rhyme is impossible to ignore.

Which brings me back to my wife–and her multiple choice question about the new, new thing.

She had listed her options. AI, of course, was on the list. So was quantum computing. Biotech convergence. Energy transition infrastructure. I walked her through each one using exactly the framework above–where is it in the capital cycle, who is doing the funding now, has the great returns window already closed.

When I finished, she looked at me with that particular expression she has–the one that tells me she was three steps ahead the whole time and said: "So the new, new thing isn't any of those–you choose ‘none of the above.’ The new, new thing is whoever figures out how to make money off all the money being thrown at it. There is nothing new about that at all. Hmm." She walked away. I knew that was just the opening discussion, she already has another set of multiple choice questions.

Indefatigable. I told you.

YESTERDAY’S MARKETS

Stocks closed higher yesterday to end the second quarter on a strong note, with the Nasdaq leading the way up 1.52%, while the S&P 500 gained 0.79% to close at 7,499 and the Dow added 136 points to finish at a record 52,319. Semiconductor names drove the session, with NVIDIA up 2.6%, AMD surging 7.7%, and Intel advancing 6% as investors looked past near-term concerns about stretched AI valuations. The 10-year Treasury yield closed at approximately 4.42%, moving higher on the day after the release of employment data. For the quarter, the S&P 500 rallied more than 14%, the Nasdaq gained roughly 20%, and the Dow added over 12%--the best quarterly performance for major US indexes since 2020.

NEXT UP

• ADP Employment Change (June) came in at 98k, lower than the expected 120k, and lower than last month’s 122k adds.

• ISM Manufacturing PMI (June) may have inched lower to 53.9 from 54.0.

• Shiny new Fed Chair Warsh will speak with other top bankers this morning in Europe. 👀