Nvidia has found a new constraint on the AI boom.

It isn’t chips. It isn’t power.

It’s money.

Alphabet, Amazon, Meta, Microsoft and Oracle are expected to spend roughly $750 billion on data centres this year. That is an extraordinary amount of capital from five of the richest companies ever assembled.

And it still may not be enough.

Some are already burning cash. Borrowing costs are rising. Shareholders are beginning to question how many billions can disappear into concrete, electricity and Nvidia chips before the returns arrive.

Nvidia’s solution is pretty direct: find more money.

This week, it recruited Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to help mobilize more than $500 billion for AI infrastructure. The financing would allow smaller cloud operators and AI companies to keep building - even when their own balance sheets cannot support it.

The world’s most valuable chipmaker is beginning to look like a car company that sells the vehicle, arranges the financing and promises the bank it can resell the car later.

There are two ways to read this.

The bullish one is absurd: demand for AI computing is so strong that even the deepest corporate pockets cannot satisfy it. Nvidia is not searching for customers → it is searching for enough capital to fund the customers already waiting.

The bearish reading is that demand increasingly requires financial engineering to continue.

Both can be true.

For TSMC (our #1 seat), the immediate equation is simple: every financed Nvidia system contains more leading-edge chips for TSMC to manufacture.

The AI boom is not running out of demand. It is running out of people rich enough to pay for it.

The Five finished its opening week down 0.35%.

SPY gained 0.48% over the identical period. QQQ gained 1.20%.

Three seats made money. Two lost it. Vertex’s decline was large enough to drag the complete portfolio below both benchmarks despite a strong opening week from Intuitive Surgical.

The Crown: Intuitive Surgical, +4.09%.
The Hot Seat: Vertex Pharmaceuticals, −6.34%.

Uber spent fifteen years convincing investors that it did not need to own the cars.

Now it plans to spend more than $10 billion acquiring them.

The company that made taxis obsolete by separating the driver from the dispatch office is preparing for a world without drivers - and apparently rediscovering the appeal of owning an enormous fleet.

This is either the final step in Uber’s transformation into the operating system for global transportation.

Or it is a very expensive identity crisis.

So what is Uber actually doing?

Uber’s great invention was not the black car summoned from an iPhone.

It was making someone else supply it.

Drivers bought the vehicles. Drivers paid for fuel, maintenance, insurance and depreciation. Uber supplied the software, matched those drivers with riders and collected a percentage of every trip.

The cars sat on other people’s balance sheets.

It took years (and billions in losses) for the model to work. But it finally does. Uber now moves more than 200 million customers, generates billions in free cash flow and can add riders without buying a corresponding number of vehicles.

That is the beauty of a marketplace. Once enough buyers and sellers gather in one place, growth becomes less expensive.

But now, robotaxis threaten to reverse that equation.

Remove the driver and the largest operating expense disappears. But so does the person who bought the car.

Someone still has to pay for the vehicle.

Uber increasingly believes that someone may have to be Uber.

Uber intends to commit more than $10 billion to autonomous transportation over the coming years. The money will fund vehicles, investments in autonomous-driving companies and the infrastructure required to put robotaxis into commercial service.

The ambition is enormous: Uber is targeting as many as 120,000 autonomous vehicles across at least 15 cities by the end of 2026. (Financial Times)

It is not betting on one winner.

Uber is working with Waymo where that relationship remains useful. It has invested in Lucid and Nuro to create a robotaxi exclusive to its platform. It has relationships with Wayve, WeRide, Waabi, Volkswagen and others across passenger vehicles, delivery and freight.

The Lucid arrangement shows how far Uber is willing to go.

Lucid supplies the electric SUV. Nuro supplies the autonomous-driving system. Uber supplies the riders, designs the passenger experience and helps finance the entire operation. At least 20,000 vehicles are expected over six years, owned either by Uber or outside fleet partners and available exclusively through Uber. Uber

Uber is not rebuilding its old self-driving laboratory. It shut that operation down in 2020 after years of losses and disappointing progress.

This time, it is buying everything around the technology.

Uber wants to become the place where autonomous vehicles become businesses.

Building a vehicle that can drive itself is only one part of operating a robotaxi network. The vehicle still needs customers, pricing, payments, routing, cleaning, charging, maintenance, regulatory support and someone to answer when a passenger leaves a phone in the back seat.

Uber already knows how to do most of that.

It also possesses something an autonomous-vehicle developer cannot manufacture in a laboratory: demand.

Uber operates across more than 70 countries and completes roughly 34 million trips each day. A robotaxi entering that marketplace does not have to wait for customers to discover a new app. It can begin earning money immediately.

That should make the vehicle more valuable on Uber than it would be alone.

The strategy is not to find the winning robotaxi and place one enormous bet. It is to back several developers, finance their expansion and make Uber the default marketplace connecting all of them with riders.

Waymo can win.

Nuro can win.

Wayve can win.

Uber wants to collect the fare either way.

There is another advantage. Human drivers do not appear everywhere Uber needs them, exactly when it needs them. Robotaxis will not immediately replace the driver network; they can fill its gaps.

Autonomous vehicles can serve quiet hours, distant neighbourhoods or routes where demand is predictable but driver supply is thin. Human drivers can absorb concerts, airport surges, bad weather and the thousand strange situations autonomous systems may initially avoid.

That hybrid network could be considerably stronger than either one alone.

A robotaxi-only service must own enough vehicles to survive its busiest hour, then watch many of them sit idle during the rest of the day. Uber can mix autonomous supply with millions of human drivers and send each trip to whichever vehicle makes the most economic sense.

The advantage is not simply having robotaxis. It is having somewhere useful to put them.

There is an inconvenient difference between operating a marketplace and owning transportation equipment.

Cars depreciate.

They crash. They require cleaning, charging, storage and repair. Sensors become obsolete. Autonomous-driving systems improve. A vehicle purchased today could become economically inferior long before its physical life ends.

Uber can bring in outside fleet owners to carry some of that burden. The company says vehicles in its Lucid-Nuro program may be owned by Uber or third-party partners.

That “or” is doing a lot of work.

If pension funds, infrastructure investors and fleet specialists ultimately finance the vehicles, Uber may preserve much of its marketplace model. It can use its balance sheet to launch the network, prove the economics and then move the assets elsewhere.

If those investors never arrive (or demand returns too slowly) Uber becomes the owner of a vast fleet of highly specialized equipment.

The partnerships create another problem.

Uber needs autonomous-vehicle developers.

It also needs those developers not to become Uber.

Waymo already has its own consumer app. Its limited Phoenix partnership with Uber ended in June, although the companies still work together in Austin and Atlanta.

If autonomous operators can attract enough riders directly, Uber becomes optional. If they cannot, Uber gains enormous bargaining power.

That is the contest hiding underneath every friendly announcement.

Uber’s strategy is more coherent than it first appears.

The company does not believe every autonomous-vehicle developer will surrender its customers to Uber. It believes the industry will produce too many vehicles, technologies and local operating challenges for every developer to build a global marketplace alone.

Uber wants to be the common layer.

Its app supplies demand. Its network balances human and autonomous vehicles. Its operating system handles the ugly work between a successful driving demonstration and a profitable ride.

The $10 billion is the price of making that future difficult to build without it.

But the investment case now contains a contradiction that cannot be waved away.

Uber is spending heavily on physical assets to protect the value of an asset-light marketplace.

If the plan works, the company will not need to invent the best robotaxi. It will own the marketplace every robotaxi needs to make money.

If it fails, Uber will have spent billions replacing independent drivers with vehicles it must finance itself.

Uber may be building the tollbooth for autonomous transportation.

First, it has to prove it isn’t buying the highway.

The author personally owns TSM, MA, VRTX, ISRG and PGR starting on the Monday (August 10th, 2026) following the first newsletter. No company discussed paid for inclusion. Personal trades in covered securities are restricted around publication.

The Five is a concentrated model portfolio published for general informational and educational purposes. It is not individualized investment, legal, accounting or tax advice and does not consider any reader’s objectives, financial circumstances or risk tolerance.

Investing involves risk, including the possible loss of principal. A five-stock portfolio is highly concentrated and should not be treated as a complete or appropriately diversified portfolio.

Model results may differ from real-world investor results because of taxes, currency conversion, bid-ask spreads, trading costs, execution availability and other factors. Past performance does not guarantee future results.

Portfolio transactions announced Sunday are recorded using the official Monday opening price under The Five Constitution. Performance figures must come from the official Ledger and are subject to correction when data errors or corporate actions are identified.