Can the AI Rally Survive 5% Treasury Yields? Nvidia, Microsoft, Meta, Amazon & Google Analysis

The AI rally has acquired an expensive rival: a US government bond. The 10-year Treasury yielded 5.28% on October 2, while the 30-year offered 5.63%. Yet the S&P 500 finished within 1% of its record. 

That puts a harder question in front of anyone tracking Nvidia, Microsoft, Meta, Amazon or Google: can their future cash returns justify today’s share prices when relatively safe dollar debt offers more than 5%? Our view is that AI can keep growing, but the investment case now needs much stronger evidence of returns on the money being spent.

Let’s break down how 5% Treasury yields test Nvidia, Microsoft, Meta, Amazon and Google.
We’ll compare valuations, AI spending and cash returns to see which investment cases have room for disappointment.

Why Do 5% Treasury Yields Matter For The AI Rally?

A Treasury yield is the return implied by a US government bond’s price and promised payments. For equity investors, it helps set the starting point for the return they demand from a riskier investment. A quoted yield is not a guaranteed one-year total return: an investor who sells before maturity can receive more or less than the bond’s face value..

One basis point equals 0.01 percentage point. These are official daily readings, which can differ from intraday market quotes.

The 10-year yield also needs to be distinguished from the Federal Reserve’s policy rate. The Fed raised its target to 3.75%-4.00% on September 16. Long-term yields reflect expectations for future short-term rates and a term premium which is the extra compensation investors demand for uncertainty over a longer holding period.

The reason yields rise matters. Stronger growth can support both higher rates and stronger company earnings. Inflation pressure or a higher premium for uncertainty can raise the return investors demand without improving business profits. Watching inflation-adjusted Treasury yields helps distinguish a higher real return from inflation compensation.

Five percent therefore changes the calculation, but does not dictate a stock-market outcome. For the wider macro background, see INDmoney’s explanation of why US Treasury yields are rising.

Why Can Stocks Rise Even When Treasury Yields Are Above 5%?

A share price depends on both the company’s earnings and the price investors are willing to pay for those earnings.

Share price = earnings per share × price-to-earnings ratio.

The price-to-earnings ratio, or P/E, tells you how much investors pay for each dollar of annual profit. At 30 times earnings, they pay $30 for $1 of earnings per share.

Higher yields can make investors less willing to pay a rich multiple. Earnings growth can still offset that pressure, but the math is less forgiving than it first appears.

Source: Author calculations. Required EPS growth = 1 ÷ (1 − multiple decline) − 1. Dividends are excluded. These scenarios do not predict how much any company’s P/E will fall.

For example, 15% earnings growth combined with a 20% lower P/E produces an 8% lower share price: 1.15 × 0.80 − 1.

That explains how an AI company can report good results while its shares disappoint. It also explains why stocks can remain resilient if operating performance keeps improving fast enough.

Our interpretation of the current market is that investors continue to assign substantial value to AI’s future earnings. The question is whether those earnings will arrive soon enough, and leave enough cash after infrastructure spending, to defend that confidence.

How Do Higher Treasury Yields Reduce The Value Of Future AI Cash Flows?

Investors value a company partly by estimating future cash flows and translating them into today’s money. The discount rate is the return required to make waiting for that money worthwhile.

When the required return rises, a fixed future payment is worth less today. A company whose value depends heavily on distant cash flows can therefore face considerable valuation pressure.

Here is a simplified illustration. Assume a business generates $10 billion of cash next year, growing by 4% annually forever. Also assume the required equity return equals the Treasury yield plus a fixed five percentage points for risk.

Source: Author model. Value = next year’s cash flow ÷ (required return − growth rate). The risk premium and growth rate are assumptions, not estimates.

Moving from the first row to the second reduces the modelled value by 16%, even though the business’s cash forecast has not changed.

Real companies have changing growth, investment needs and risk. The lesson is narrower: a strong business can become a less rewarding investment if investors originally paid a price that assumed a lower required return.

Cash-rich companies also face this opportunity cost. They may fund a project without borrowing, but shareholders can still ask whether using that cash on AI creates enough value.

NVDA, MSFT, META, AMZN and GOOGL: What Do Their Valuations Show?

The valuation picture contains a surprise. Nvidia’s forward P/E is the lowest in this snapshot, while Alphabet’s trailing P/E is much lower than its forward P/E.

Trailing P/E uses reported earnings from the latest 12 months. Forward P/E uses forecast earnings, so it changes when analysts revise their estimates.

Valuation snapshot accessed October 5 before the US market opened, using October 2 closing prices. P/E data: Stock Analysis. *FCF yields are author calculations using company cash-flow definitions and rounded market values.

Free-cash-flow yield divides the annual cash left after capital spending by the company’s market value. It is not a dividend yield or a promised investor return. The underlying cash calculations appear in the next comparison.

These low cash yields show that investors are paying for future growth and eventual cash recovery. They do not, by themselves, establish that a Treasury is the better investment.

Nvidia’s 19.4× forward P/E also deserves scrutiny. If forecast earnings were cut by 20% while the share price stayed unchanged, that multiple would become about 24.3×. A modest-looking forward valuation can depend on demanding earnings expectations.

There is another reason to be careful with the apparently lower multiples at Amazon and Alphabet.

Why Can Headline P/E Ratios Give A Misleading Picture Of AI Profits?

A company can report a large profit because the value of an investment increased. That gain can improve EPS without bringing in cash from customers.

Alphabet’s Q2 2026 release makes the distinction unusually clear. Its reported diluted EPS was $9.11. The company disclosed that net gains on equity securities contributed $6.26 per share after tax. Subtracting that contribution leaves $2.85. This removes that specific item; it is not a complete adjustment for every unusual expense or gain.

Amazon reported $62.6 billion of Q2 net income, while disclosing $53.4 billion of non-operating pre-tax other income, primarily from its Anthropic investments. The pre-tax gain cannot simply be subtracted from after-tax net income to calculate an adjusted profit.

These investment interests can have real economic value. But a valuation gain has a different role from recurring cloud, advertising or software cash receipts.

For this article’s question, the useful test is whether recurring operating profits and cash collections can fund the AI buildout. A lower P/E caused by an investment revaluation gives us limited help with that assessment.

Which AI Companies Can Fund Their Capital Spending From Operating Cash?

Operating cash flow is the cash generated by a company’s operations. Free cash flow generally subtracts capital spending, such as purchases of servers and data-centre equipment.

Comparing that spending with operating cash shows how much of the current cash generation is being reinvested.

Figures cover the trailing 12 months to June 30, 2026, except Nvidia, which ends July 26. Nvidia and Meta trailing totals are calculated as the preceding financial year plus the latest first half, minus the prior first half. Microsoft’s figures cover its full fiscal year. 

Sources: company earnings releases.

*Definitions differ. Nvidia includes intangible-asset spending and relevant principal payments; Meta includes finance-lease principal. Amazon nets property proceeds and incentives. Microsoft’s cash measure and Alphabet’s measure exclude finance-lease principal. These are company-wide figures, not separately disclosed AI budgets, and exclude financial investments and acquisitions.

The table measures cash available after those specified deductions. It does not capture every funding commitment or establish whether the spending earns an adequate return. Still, Amazon spent more than its operations generated, while Nvidia’s own capital spending consumed a much smaller share.

Nvidia: strong supplier cash flow, with customer funding still important

Nvidia collects revenue when customers purchase its infrastructure. Those customers must then earn a return from using it. Think of a car dealer supplying a taxi fleet: the dealer’s economics and the fleet operator’s economics are connected, but their cash arrives at different times.

Nvidia’s Q2 Data Center revenue reached $89.0 billion, up 117% from a year earlier. That is substantial commercial demand today,

However, its low capital-spending ratio should not be mistaken for an absence of cash commitments. In the first half of FY2027, Nvidia spent $42.4 billion purchasing equity securities. Its cash-flow statement also recorded a $24.6 billion use of cash from growth in accounts receivable, the money customers owe for sales already recorded. These items merit attention alongside hardware demand.

Our view: Nvidia has the strongest reported cash retention in this group after the specified capital deductions. The investment case still depends on customers sustaining orders, paying their bills and earning sufficient returns on installed capacity.

Microsoft: the clearest cash cushion among the four large spenders

Microsoft retained approximately $67 billion after cash purchases of property and equipment over the latest year. Its 63.4% spending ratio is lower than the other three large infrastructure spenders in this comparison.

There is also evidence of monetisation. Azure and other cloud services revenue grew 43% in Q4 FY2026, and management reported more than 30 million paid Microsoft 365 Copilot seats, or subscription licences. These are different revenue routes through which AI can generate returns.

The lease distinction matters. Microsoft reported $41 billion of quarterly capital expenditure, compared with $35.8 billion paid in cash for property and equipment.

We see comparatively stronger funding flexibility here. Investors still need to examine cash after investment and lease obligations, alongside the growth in cloud and software demand.

Meta: an existing advertising business, but a thinner recent cash buffer

Meta can benefit from AI through advertising recommendations and tools, so returns need not come only from charging users for a new AI product.

Its Q2 ad impressions rose 14%, while the average price per ad increased 12%. Those figures show growth in the existing advertising business; they do not isolate how much AI contributed.

The recent cash position deserves closer attention. Q2 operating cash flow was $31.86 billion and capital spending, including lease principal, was $31.08 billion. Just $784 million remained, approximately 2.5% of operating cash.

One quarter does not establish a permanent funding problem. But it makes the next few quarters’ spending, advertising returns and cash conversion particularly important. Meta’s AI ambitions need to translate into enough additional cash to restore a wider buffer.

Amazon: strong AWS growth alongside the toughest current funding test

Amazon presents the sharpest tension in the table. Its trailing capital spending exceeded operating cash flow by $7.6 billion, even as AWS quarterly revenue grew 37% and AWS operating income reached $16.6 billion.

That combination shows why strong cloud demand and funding pressure can exist together. Amazon’s cash figures also include retail and other businesses, so the spending ratio is not an AWS return measure.

Our view: Amazon’s current cash profile requires particularly convincing evidence that new capacity will produce enough future cash. Higher rates make delays and weak utilisation harder to absorb. Investors should watch when growth in operating cash begins to catch up with infrastructure spending.

Alphabet: cloud momentum must support a much larger infrastructure bill

Google Cloud revenue grew 82% in Q2, reaching $24.8 billion, while its operating margin reached 35.6%. That means approximately $35.60 in operating profit per $100 of revenue, providing evidence of current monetisation.

But Alphabet’s Q2 capital spending of $44.9 billion exceeded its $39.1 billion operating cash flow. Management also raised its 2026 capital-spending forecast to $195 billion-$205 billion and warned of higher depreciation and data-centre operating costs.

We would judge Alphabet through recurring operating earnings and cash recovery across its businesses. Its unusually low trailing P/E gives an incomplete picture of that challenge. For the underlying revenue and profit mix, see how Google makes money.

Can AI Projects Earn Enough To Clear A Higher Return Hurdle?

Capital expenditure creates two separate tests. First, cash leaves when equipment is purchased. Second, depreciation allocates the equipment’s cost across its estimated useful life, affecting reported profits over time.

Microsoft said roughly two-thirds of its Q4 capital expenditure went into shorter-lived assets, primarily CPUs and GPUs. The earnings challenge therefore includes equipment economics as well as the cost of financing buildings.

Hardware life should not be guessed from a product launch. Nasdaq’s analysis noted continuing rental demand for older Nvidia GPUs. A newer chip does not automatically make the previous generation commercially worthless.

A simple project model helps separate these risks. Assume $100 billion is spent today on equipment that produces $30 billion of annual net cash for five years, with no resale value. Cash is after operating costs, maintenance and taxes. There is no additional growth investment. The five-year life is an illustration, not an estimate for a particular GPU.

Source: Author discounted-cash-flow calculations. Payments arrive at year-end. The final column is net present value, or NPV, which measures value created above the required return. These project hurdle rates are assumptions, not estimates of the companies’ actual financing costs.

The rate increase alone reduces value but leaves this hypothetical project viable. Lower cash receipts or a delay remove much more of its original cushion. The cash shortfall would also hurt at the lower rate; higher rates make the weakness harder to absorb.

At the 10.25% hurdle, the project needs approximately $26.5 billion annually to break even in present-value terms. That cash includes recovery of the initial investment. It is not a 26.5% profit margin or an annual investment return.

This is why we think utilisation, pricing and equipment life deserve as much attention as Treasury yields. Utilisation means how much available computing capacity is doing paid work. Faster adoption only creates attractive returns if receipts adequately cover operating costs and the capital invested.

What Could Put Pressure On The AI Rally First?

Our view is that investors should look for weakening cash economics before waiting for a dramatic slowdown in AI usage.

The following signals connect the bond-market story to company results.

These are questions for further analysis, not automatic failure signals. Contract timing, capacity coming online and working-capital movements can explain temporary changes.

A particularly useful question is who ultimately funds the AI customer. A profitable enterprise paying for productivity improvements has a different funding profile from a business that needs repeated capital raises to pay its computing bills. INDmoney’s analysis of OpenAI’s cash burn and the implications for AI stocks explores these connections.

Holding all five stocks also creates exposure to a shared spending cycle. Nvidia benefits from infrastructure demand; several of the other companies finance that demand. Their different business models do not remove the connection between customer returns and future hardware orders.

How Should Investors Read The Next Move In Treasury Yields?

The useful comparison combines yields with business cash performance.

These are conditional scenarios, not stock-price forecasts. Falling yields help most when the underlying earnings outlook remains healthy.

We would give more weight to growth in recurring cash per share, the pace of capital spending and evidence that installed capacity earns an adequate return. We would require a wider margin for error where the valuation depends on a rapid cash recovery or where reported profits include large investment gains.

The project model shows why 5% yields alone need not end the AI rally. It also shows how quickly an attractive investment can lose value when cash arrives late or falls short. The next results need to establish how much recurring cash these companies can retain after funding the infrastructure behind their growth.