Want to Invest in AI? These Stocks Offer Very Different Risk

By AltIndex Research · 11 min read · August 24, 6:12 am

AI demand is still booming. But the companies benefiting from it are taking very different financial risks. Some are generating enormous profits selling the chips behind the boom. Others are borrowing billions to build the data centers that house them. For investors, that distinction is becoming increasingly important.

The AI investment thesis has been relatively straightforward for the past few years: demand for computing power is exploding, hyperscalers are spending hundreds of billions of dollars, and the companies supplying that infrastructure are benefiting.

But there is another side to the boom that investors increasingly need to consider: someone has to pay for all those data centers.

That matters more now because financing conditions have become less forgiving. AI-related companies have flooded debt markets to fund infrastructure expansion, while investors have started demanding higher yields to finance some of those projects.

For retail investors looking for exposure to AI, that creates two very different types of investment.

On one side are companies such as CoreWeave (CRWV), Nebius (NBIS), Applied Digital (APLD) and TeraWulf (WULF). They are building the infrastructure needed to satisfy enormous AI demand, often using significant amounts of outside capital.

On the other are companies such as Nvidia (NVDA), TSMC (TSM), Micron (MU), ASML (ASML) and Marvell (MRVL). They sell the chips, memory and manufacturing technology going into that infrastructure, but generally don't need to finance the customer's data center themselves.

Both groups can benefit enormously if AI demand continues growing.

The risks, however, are very different.

CoreWeave Shows Both Sides of the AI Boom

CoreWeave (CRWV) may be the clearest example.

There is little evidence that demand is the problem.

CoreWeave generated $2.58 billion of revenue in the second quarter, up 112% from a year earlier, and finished June with approximately $104 billion of revenue backlog. It also added another $25 billion of customer commitments early in the third quarter.

But delivering all that computing power requires staggering investment.

CoreWeave expects $35 billion to $39 billion of capital spending in 2026. Its second-quarter interest expense alone reached approximately $640 million, up from $267 million a year earlier.

That's the trade-off.

CoreWeave has secured extraordinary demand. But it needs enormous amounts of capital to fulfill that demand.

Our alternative data supports the growth side of the story. AltIndex estimates CoreWeave's LinkedIn headcount has increased approximately 40% over the past six months, while estimated website traffic has risen about 35%.

Its AI Score, however, currently sits at just 56 out of 100. The customer component is much stronger, at 100, illustrating an interesting divide: our data sees extremely strong customer signals, but the broader picture is considerably less bullish.

For investors, CoreWeave is therefore not simply a bet on whether AI demand continues.

It's also a bet that the economics of building and financing that capacity remain attractive enough to justify the extraordinary investment required.

Nebius Is Taking Risk — but Its Growth Is Exceptional

Nebius (NBIS) offers a different version of the same trade.

The company recently raised $5 billion through convertible bonds, bringing its total convertible debt to approximately $12 billion. The financing will help fund its rapidly expanding data-center footprint.

Nebius spent roughly $5.7 billion on capital expenditures in Q2. Yet revenue is exploding: Q2 sales reached approximately $582 million, roughly five times the year-earlier level.

This is exactly where alternative data becomes useful.

An AI Score of 81 is the highest among the 12 AI stocks we analyzed.

So Nebius isn't a case where debt is rising while the underlying business stagnates. Quite the opposite.

The company is taking substantial financial risk while almost every growth signal we track is accelerating.

That doesn't remove the risk. It explains what investors are receiving in exchange for taking it.

TeraWulf Shows Why Investors Need to Look Beyond the AI Story

TeraWulf (WULF) is a useful contrast.

The company's transition toward AI infrastructure is progressing quickly. Q2 revenue reached $44.8 million, with $31.9 million—or about 71%—coming from HPC leasing. It also finished the quarter with approximately $3 billion of cash and restricted cash.

Our employment data is bullish as well. Estimated LinkedIn headcount has increased approximately 73% in six months, while TeraWulf's AI Employment Score has reached 100.

Its overall AI Score is a relatively strong 73.

But the financial structure deserves attention. The company's filings show sharply higher expenses as it builds its HPC operations, while its balance sheet has expanded alongside those projects.

That's an important distinction for investors.

Hiring tells us TeraWulf is preparing for growth.

It doesn't tell us that the eventual return on billions of dollars of infrastructure investment will be attractive.

We excluded TeraWulf's employee-review signals from this analysis because the available review sample is not sufficiently representative.

Now Compare That With Nvidia

Nvidia (NVDA) participates in the same AI boom from almost the opposite position.

CoreWeave raises capital partly to buy computing infrastructure.

Nvidia sells the chips.

Nvidia's most recently reported quarter produced $81.6 billion in revenue, up 20% sequentially, following a fiscal year in which its profitability and cash generation reached extraordinary levels.

Our alternative data remains supportive as well.

Nvidia currently carries an AI Score of 75, while our LinkedIn estimates show its workforce roughly 13% larger than six months ago.

The crucial difference is that Nvidia does not need to borrow tens of billions of dollars every year to build data centers for its customers.

Instead, many of those customers and infrastructure providers are raising the money—and then spending part of it on Nvidia hardware.

In a world where AI spending keeps increasing but financing becomes more expensive, that distinction can become extremely valuable.

TSMC Offers a Similar Trade-Off

Taiwan Semiconductor Manufacturing (TSM) is capital-intensive itself. Semiconductor fabs are among the most expensive factories on Earth.

But TSMC is also extraordinarily profitable.

Second-quarter revenue reached $40.2 billion, with a 67.7% gross margin and 60.3% operating margin. Management expects third-quarter revenue of $44.6 billion to $45.8 billion.

AltIndex gives TSM an AI Score of 75.

Our alternative data also shows job postings up approximately 42% over six months, LinkedIn headcount up roughly 8%, and estimated website traffic up approximately 26%.

That's an unusually attractive combination:

AI demand + strong growth signals + enormous profitability.

TSMC obviously carries risks of its own—most notably geopolitical risk surrounding Taiwan—but dependence on speculative external financing isn't one of the central ones.

Micron Shows How Profitable the Memory Boom Has Become

Micron (MU) provides another example.

Its latest reported quarter produced $41.46 billion in revenue, compared with $9.30 billion a year earlier. Operating cash flow reached $25.39 billion, versus $4.61 billion in the year-earlier quarter.

Those numbers are extraordinary and reflect the enormous demand for memory used in AI infrastructure.

AltIndex estimates Micron's headcount has increased approximately 15% over six months, while web traffic is up roughly 32%.

Interestingly, its AI Score is only 60.

That is worth paying attention to. Strong financial results do not automatically produce a high AltIndex AI Score because the model considers a broader collection of fundamental, employment, customer and audience signals.

For investors, Micron therefore looks financially much less dependent on external capital than the neocloud operators—but our broader signals aren't currently as bullish as the headline earnings numbers might suggest.

ASML and Marvell Offer Two More Ways to Sell Into the Boom

ASML (ASML) sits even further upstream.

Its lithography systems are required to manufacture the world's most advanced chips. Q2 sales reached €9.3 billion, with a 54% gross margin and €2.9 billion of net income. AI-driven demand has become strong enough that ASML raised its 2026 sales outlook to €43 billion to €45 billion.

Marvell (MRVL), meanwhile, is becoming increasingly important in custom AI silicon and networking.

Its most recently reported quarter generated $2.42 billion of revenue and a record $639 million of operating cash flow. More importantly, Google has now struck a major custom-chip agreement with Marvell that could result in up to $120 billion of purchases through fiscal 2033 if the associated targets are reached.

Marvell currently carries an AI Score of 72.

That puts both companies firmly in the category of businesses selling critical components into the AI buildout rather than primarily financing the data centers consuming them.

Not Every Data-Center Stock Looks the Same

This is also why we'd avoid treating every AI infrastructure operator as equally risky.

Applied Digital (APLD), for example, finished February with approximately $2.1 billion of cash, cash equivalents and restricted cash against $2.7 billion of debt. Its fiscal fourth-quarter adjusted revenue subsequently reached $240.4 million, compared with $38 million a year earlier.

Our data shows Applied Digital's headcount up approximately 34% over six months.

IREN (IREN) also maintains substantial liquidity while pursuing an aggressive expansion strategy. It reported approximately $2.6 billion of cash as of April 30, while its AI infrastructure ambitions have expanded rapidly.

Cipher Digital (CIFR), meanwhile, remains in transition. Q2 bitcoin-mining revenue declined to $24.8 million from $43.6 million a year earlier, even as the company invests toward its broader data-center strategy.

Our estimates show Cipher's LinkedIn headcount up approximately 48% over six months and job postings up around 67%.

For Applied Digital and Cipher, we excluded employee-review data because the available samples aren't sufficiently representative.

These aren't necessarily companies to avoid.

They are simply investments where execution, financing and utilization matter much more.

The AI Trade Is Becoming a Risk-Reward Spectrum

For investors, we'd think about the current AI landscape roughly like this:

Broadcom (AVGO) belongs broadly in the supplier group as well, although the distinction is becoming less clean. Broadcom generated a record $22.2 billion of revenue in fiscal Q2, up 48%, with adjusted EBITDA of $15.2 billion. But it is also participating in financing structures around enormous AI infrastructure projects, including reported discussions involving special-purpose vehicles rather than simply putting all the associated debt directly onto Broadcom's corporate balance sheet.

So "supplier" does not automatically mean "low risk."

And "data-center operator" does not automatically mean "bad investment."

The important question is what risk you're being paid to take.

Where Would We Put Our Money?

For a retail investor who believes AI infrastructure spending will continue growing, the current data argues for separating AI demand risk from AI financing risk.

Nvidia, TSMC, Micron, ASML and Marvell provide exposure to AI infrastructure spending without requiring investors to make the same bet on cheap and continuously available external financing.

That's the cleaner side of the trade.

CoreWeave, Nebius and the emerging data-center operators potentially offer much greater operating leverage. If AI compute demand stays strong, capacity remains highly utilized and financing remains available, their revenue can grow extraordinarily quickly.

Nebius demonstrates that potential particularly well: revenue up roughly 300%, jobs up 254%, headcount up 72%, web traffic up 62%, and an AI Score of 81.

But those returns come with another variable that Nvidia investors don't have to worry about nearly as much:

Can the company keep financing the buildout on attractive terms?

That's increasingly important because the AI boom isn't running out of demand.

It is running into the enormous cost of supplying it.

For investors deciding where to put money in the AI trade, that may become one of the most important distinctions of the next several years.