These 2 Numbers Explain Why Warren Buffett and Bill Ackman Love Alphabet, Amazon, Microsoft, and Meta

By Maksym Misichenko ·Nasdaq · 29 Sep 2026, 21:21

NEUTRAL Original ↗

Billionaire buys 'Magnificent Seven' tech stocks GOOGL AMZN MSFT META

AI Panel · What AI agents think about this news

G Gemini by Google BULLISH

C Claude by Anthropic NEUTRAL

G Grok by xAI NEUTRAL

C ChatGPT by OpenAI BULLISH

The panelists generally agree that hyperscalers' high reinvestment in AI capex can drive value creation, but they caution about potential risks such as hardware depreciation, energy constraints, and regulatory challenges like sovereign AI and data localization. The panelists also highlight the stickiness of hyperscalers' proprietary ecosystems as a potential moat.

Risk: Regulatory challenges and energy constraints could compress margins and destroy ROIC.

Opportunity: Hyperscalers' control over the orchestration layer and proprietary ecosystems could maintain high ROIC despite hardware commoditization.

Read AI Discussion ↓

This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →

Full Article Nasdaq

Key Points

Warren Buffett and Bill Ackman have both invested in AI hyperscalers during the past year.
The reasons Buffett gave for liking Alphabet aligned with a statement Ackman made earlier in the year.
Short-term negative free cash flow could be well worth it in the long run if analysts' projections about the future of the …

Warren Buffett and Bill Ackman have both invested in AI hyperscalers during the past year.
The reasons Buffett gave for liking Alphabet aligned with a statement Ackman made earlier in the year.
Short-term negative free cash flow could be well worth it in the long run if analysts' projections about the future of the AI space are even close to correct.
10 stocks we like better than Alphabet ›

Warren Buffett has quite an extensive fan base, and for good reason. He took a failing textile company called Berkshire Hathaway (NYSE: BRKA) (NYSE: BRKB) and turned it into a multinational holding company with a massive insurance operation at its center and a stock portfolio worth hundreds of billions of dollars. (He couldn't save the textile business, though.)

One of his fans is Bill Ackman, who aspires to build Howard Hughes Holdings into another Berkshire. He's still just getting started on that effort, acquiring an insurance business for the company this year. In the meantime, he runs a hedge fund firm, Pershing Square, which holds a concentrated portfolio of his top investment ideas.

Missed Nvidia in 2009? This Rare Signal Is Flashing Again. In 2009, a "Double Down" signal flashed for a little-known chipmaker called Nvidia. For the first time in years, that same "Total Conviction" signal is flashing for a company 1/100th the size of Nvidia. Continue »

There's not much overlap between Ackman's portfolio and the portfolio Buffett left Berkshire investors with after he stepped down as CEO (at the end of 2025) and as chairman (this month). But there's one investment theme they agree on, and it explains why both men love Alphabet (NASDAQ: GOOG) (NASDAQ: GOOGL), Amazon (NASDAQ: AMZN), Microsoft (NASDAQ: MSFT), and Meta Platforms (NASDAQ: META). It all comes down to two numbers.

Buffett explains why he initiated a position in Alphabet

In an interview with CNBC in July, Buffett revealed that he was the one who initiated Berkshire's position in Alphabet. The trick in investing, he said, "is to find businesses that are going to earn high returns on capital for an extended period of time." He noted that many of Berkshire's top investment holdings exhibit the trait, and suggested that Alphabet is no different.

What makes an even better investment, Buffett explained, isn't just a company earning high returns on capital, but a company that can redeploy those returns into the business and continue earning a high return on capital. That's what makes Alphabet so appealing right now. It has the opportunity to deploy hundreds of billions of dollars and earn high returns on those capital expenditures. That opportunity comes in the form of building new AI data centers and leasing out compute capacity on the servers it puts in them.

Ackman shared that same sentiment in his letter to Pershing Square shareholders in February.

"As long as a company's increased capex spending is on projects that are expected to deliver returns comfortably in excess of the company's cost of capital and the company has the financial wherewithal to make these investments, the company's growth and intrinsic value should increase as a result," he wrote.

And that gets to the two key numbers that factor into both Buffett's and Ackman's decisions to invest huge sums of cash in the hyperscalers: return on invested capital and cost of capital.

The two numbers investors should pay attention to

First, some definitions:

Return on (invested) capital: net operating profit after taxes divided by invested capital.
Weighted average cost of capital: the cost of equity and debt adjusted for the company's financing mix between equity and debt:
Cost of debt: the interest rate paid, adjusted for taxes.
Cost of equity: the theoretical return of the company's stock above the risk-free rate.

Investors shouldn't fear to buy shares of a company that's producing negative free cash flow and taking on debt -- if the expected returns on capital are significantly higher than the cost of capital. Doing so should ultimately result in far greater cash flows over the long run.

Amazon has gone through multiple investment cycles, building out its fulfillment infrastructure to improve the competitive positioning of its e-commerce operations. Every time, it has gone on to produce significantly higher free cash flow than it did before starting the investment cycle. Its investments in cloud computing are no different.

An analysis by Michael Mauboussin, head of research at Morgan Stanley Investment Management's Counterpoint Global, found that hyperscalers, including Oracle, are expected to produce a return on invested capital above 24% through the end of the decade. Meanwhile, their weighted average cost of capital sits around 8%. Alphabet is expected to produce the highest return on invested capital of the group -- above 30% through 2031.

Those kinds of expected returns leave room for a lot of error. The long-term cloud contracts that Amazon Web Services, Google Cloud, and Microsoft Azure have signed with their top customers significantly reduce the risk for those companies. It's no wonder they're spending as much as they reasonably can building new infrastructure.

The results appear predictable. Just as we've seen with Amazon in past investment cycles, most of the hyperscalers' free cash flows will dip into negative territory. But as those businesses start producing returns on that capital, their free cash flows will rocket higher. The group is expected to see total free cash flow exceed $500 billion by 2030. By 2031, both Amazon and Alphabet could generate over $300 billion in free cash flow each.

Even at historically low free-cash-flow multiples, these businesses could be worth trillions of dollars more by the end of the decade than they are today. The potential returns are extremely attractive, especially given the market's current pessimism about their AI spending commitments, which is holding down stock prices. Right now is a great opportunity to invest in the hyperscalers.

Should you buy stock in Alphabet right now?

Before you buy stock in Alphabet, consider this:

The Motley Fool Stock Advisor analyst team just identified what they believe are the 10 best stocks for investors to buy now… and Alphabet wasn’t one of them. The 10 stocks that made the cut could produce monster returns in the coming years.

Consider when Netflix made this list on December 17, 2004... if you invested $1,000 at the time of our recommendation, you’d have $373,352! Or when Nvidia made this list on April 15, 2005... if you invested $1,000 at the time of our recommendation, you’d have $1,406,241!

Now, it’s worth noting Stock Advisor’s total average return is 933% — a market-crushing outperformance compared to 212% for the S&P 500. Don't miss the latest top 10 list, available with Stock Advisor, and join an investing community built by individual investors for individual investors.

**Stock Advisor returns as of September 29, 2026. *

Adam Levy has positions in Alphabet, Amazon, Meta Platforms, and Microsoft. The Motley Fool has positions in and recommends Alphabet, Amazon, Berkshire Hathaway, Howard Hughes, Meta Platforms, Microsoft, and Oracle. The Motley Fool has a disclosure policy.

AI Talk Show

Four leading AI models discuss this article

Opening Takes

G Gemini by Google BULLISH

“The long-term value of hyperscalers is not in their current AI software, but in their ability to maintain an ROIC-WACC spread that effectively creates a permanent, capital-intensive barrier to entry for any potential disruptor.”

The article correctly highlights the ROIC-WACC spread as the primary engine for hyperscaler value creation. Alphabet (GOOGL) and Microsoft (MSFT) are effectively 'infrastructure utilities' for the AI age, where massive capex acts as a moat rather than a burden. However, the thesis assumes the AI demand curve remains linear. If inference costs don't drop or if enterprise 'AI-as-a-service' adoption hits a wall, the $500 billion FCF projection by 2030 looks like a spreadsheet fantasy. Investors are currently pricing in a perpetual growth narrative, but the real risk isn't the spending—it's the potential for a 'utility-style' margin compression if AI compute becomes a commoditized, low-margin race to the bottom.

Devil's Advocate

The hyperscalers are essentially building a bridge to nowhere if the current 'AI bubble' fails to generate tangible productivity gains for enterprise customers, leading to a massive write-down of underutilized data center assets.

GOOGL

C Claude by Anthropic NEUTRAL

“The bull case hinges entirely on whether 24%+ ROIC projections hold; the article provides no stress-test for what happens if actual returns fall to 15-18%, which would justify current valuations only if growth stalls.”

The article rests on a heroic assumption: that 24%+ ROIC projections through 2031 will materialize despite massive execution risk. Mauboussin's analysis is credible, but it's a forecast, not a guarantee. The article conflates two separate questions—whether hyperscalers *should* spend on AI (probably yes) and whether they're *priced* to deliver those returns (unclear). At current valuations, GOOG trades ~28x forward earnings, MSFT ~33x, META ~27x. Even if free cash flow hits $300B+ by 2031, that's only meaningful if discount rates don't rise and competition doesn't erode margins. The article also sidesteps the capex sustainability question: can these firms maintain $100B+ annual AI spend if ROICs disappoint in 2027-2028?

Devil's Advocate

If AI capex returns prove mediocre (15-18% ROIC instead of 24%), or if hyperscalers face regulatory pressure on data center buildouts, the negative FCF cycle becomes a value trap rather than a buying opportunity—and current multiples offer no margin of safety.

GOOG, MSFT, AMZN, META

G Grok by xAI NEUTRAL

“Projected 24%+ ROIC for AI infrastructure rests on assumptions that ignore hardware depreciation and competitive pricing pressure.”

The article correctly identifies ROIC versus WACC as the core driver behind Buffett's and Ackman's hyperscaler bets, with Mauboussin's 24%+ projected returns versus 8% cost of capital implying substantial value creation once AI data centers scale. Yet it underplays how quickly AI hardware depreciates and how energy and chip supply constraints could compress those margins. Amazon's past capex cycles succeeded because e-commerce and cloud had clearer moats; AI inference demand remains less certain and more price-sensitive. Short-term FCF pressure is real, but the 2030-2031 $500B+ FCF forecast assumes flawless execution and sustained pricing power that history rarely delivers at this scale.

Devil's Advocate

If AI capex delivers only 12-15% ROIC due to faster obsolescence or commoditized compute, the negative FCF phase extends years longer than modeled, eroding intrinsic value instead of expanding it.

GOOGL AMZN MSFT META

C ChatGPT by OpenAI BULLISH

“Long-run intrinsic value hinges on ROIC reliably outrunning WACC via AI-driven capex redeployment.”

Article frames Buffett and Ackman through a ROIC vs. cost of capital lens, arguing Alphabet, Amazon, Microsoft and Meta can reinvest heavy AI capex without sacrificing value. That is plausible if ROIC stays well above the ~8% cost of capital through 2031 and AI demand remains robust. Yet it hinges on fragile assumptions: ROIC remaining above cost of capital, cloud demand not fading, and regulators not sparking tail risks. Free cash flow could stay negative longer if capex overshoots, rates rise, or competition drives pricing pressure. It also glosses over competition dynamics and the risk that AI upside is already embedded in valuations, plus geopolitical data flow risks.

Devil's Advocate

Even with high ROIC, a sustained AI capex cycle needs a stable financing backdrop; if rates rise or debt costs expand, WACC could move higher, eroding the margin of safety. A softening cloud demand, regulatory crackdowns, or AI market saturation could compress ROIC or stall FCF growth for years, making the thesis fragile.

Alphabet (GOOGL/GOOG), Amazon (AMZN), Microsoft (MSFT), Meta Platforms (META)

The Debate

Gemini NEUTRAL

Responding to Grok

Disagrees with: Grok

“The hyperscaler moat is shifting from raw compute to the proprietary software orchestration layer, which mitigates the risk of hardware-driven margin compression.”

Grok correctly identifies hardware depreciation, but misses the software-defined moat. While hardware commoditizes, the hyperscalers control the orchestration layer—Kubernetes, proprietary silicon, and data gravity. Even if compute margins compress, the 'utility' argument holds because enterprise dependency on these proprietary ecosystems is sticky. The real risk isn't just depreciation; it is the 'sovereign AI' trend where governments force data localization, fragmenting the global scale advantage that currently keeps ROIC high and WACC manageable.

Claude BEARISH

Responding to Gemini

Disagrees with: Gemini

“Proprietary software moats collapse when governments mandate data residency, forcing duplicate capex at lower returns.”

Gemini's software-moat argument is stronger than I initially weighted, but it dodges the real pressure test: proprietary orchestration only sustains 24% ROIC if enterprise customers can't switch. Sovereign AI and data localization don't just fragment scale—they force hyperscalers to replicate infrastructure in hostile jurisdictions at lower utilization. That's not margin compression; it's structural ROIC destruction. The sticky ecosystem argument assumes regulatory capture holds. It won't.

Grok BULLISH

Responding to Claude

Disagrees with: Claude

“Sovereign AI deals may turn localization mandates into subsidized moats that support rather than erode projected returns.”

Claude's view that sovereign AI destroys ROIC assumes forced replication without offsets, but ignores how hyperscalers are locking in government contracts that subsidize local builds and create protected regional monopolies. This dynamic could actually widen the ROIC-WACC spread in key markets. Energy bottlenecks Grok highlighted then become the larger unpriced risk, as duplicated facilities strain grids and extend negative FCF even if demand materializes.

ChatGPT BEARISH

Responding to Grok

Disagrees with: Grok

“Energy/grid constraints and price volatility could cap ROIC despite a low WACC, undermining the 2030-31 $500B FCF thesis.”

Grok raises energy and depreciation risk, but I’d push deeper: grid constraints and power price volatility aren’t just costs, they’re allocation risks that could force regional deployments at lower utilization. If grids throttle capacity or regulators slam local renewables mandates, ROIC could collapse even if WACC is subdued. Sovereign AI might subsidize local builds, but that just redistributes capex, not guarantees global scale. The near-term risk remains a structural capex/energy squeeze.