The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Sometimes—but a pullback alone is not a buy signal. An AI-exposed stock is worth buying only when its future earnings and cash generation can plausibly justify the current share price, and when you can tolerate the risks, time horizon and volatility involved. The same “AI” label covers chip suppliers, cloud platforms and software companies with very different economics.
Recent results show extraordinary demand alongside rising infrastructure costs. NVIDIA reported fiscal second-quarter 2027 revenue of $96.2 billion, while Microsoft and Amazon described strong cloud and AI growth. None of those figures, by themselves, proves that any stock is cheap after a decline. You still need current price data, updated filings and a valuation that fits realistic assumptions.
As an Amazon Associate I earn from qualifying purchases.
What a pullback tells you—and what it does not
A pullback is a decline from a recent high, but the term does not identify the cause, size or investment opportunity. A stock can fall because short-term traders take profits, because expected growth is being revised lower, because interest rates change the value of future cash flows, or because investors discover a business problem. Those situations require different decisions.
The key distinction is between price movement and business value. If a company’s expected earnings and cash flows are unchanged while the share price falls, the valuation may become more attractive. If the decline reflects weaker demand, lower margins, delayed capacity or a permanently higher cost base, the lower price may simply reflect lower value.
#1 Best Overall
Do not infer the size of the latest pullback from this article. The available company disclosures do not establish today’s share prices, current drawdowns or comparable forward valuation ratios. Obtain those figures from a current market-data source and the latest regulatory filings before making a stock-specific decision.
“AI stocks” are several different businesses
Comparing every AI-labeled company as if it were one sector hides the risks that matter most. Start by identifying how a company earns money from AI and which costs it must bear to deliver that revenue.
| Business role | How AI demand can help | Costs and risks to examine |
|---|---|---|
| AI computing and infrastructure supplier | Sells processors, networking, systems or related equipment to data-center operators. | Customer concentration, product cycles, supply constraints, competition, export rules and whether customers can fund enough data-center capacity. |
| Cloud platform | Charges for computing, storage, model services and applications as customers run more workloads. | Power and data-center spending, utilization, pricing pressure, depreciation, energy availability and the margin impact of serving AI workloads. |
| Software or service provider | Adds AI features to existing products or sells new AI-powered services. | Ability to charge for features, inference costs, customer retention, intellectual-property exposure and whether usage converts into recurring cash flow. |
NVIDIA is primarily an AI-computing infrastructure supplier. Microsoft combines cloud infrastructure with software and services. Amazon combines AWS cloud operations with retail, advertising and other businesses. Their reported growth rates cannot be compared without adjusting for those different models.
What the latest disclosures actually show
The following figures are reported or management-provided information from 2026. They describe operating momentum and investment requirements; they are not valuation conclusions.
| Company and period | Reported evidence | Important qualification |
|---|---|---|
| NVIDIA, fiscal Q2 2027 (August 26, 2026 release) | Revenue of $96.2 billion, up 106% year over year; Data Center revenue of $89.0 billion, up 117% year over year. | Quarterly results do not establish that the shares are attractively priced. Future sales also depend on customers’ ability to secure land, power, data-center shells and capital. |
| Microsoft, FY2026 Q3 | Total revenue increased 18% year over year. Microsoft Cloud gross margin percentage was 66%. | The cloud gross margin percentage decreased as Microsoft continued investing in AI infrastructure and supported growing AI product usage, partly offset by efficiency gains. |
| Microsoft, FY2026 Q4 call | Management said calendar-year 2026 capital-expenditure expectations were approximately $175 billion after a change in lease mix and expected FY2027 capital expenditure to grow year over year. | This is management’s outlook at the time of the call, not realized spending or a guarantee of future growth. |
| Amazon, Q2 2026 results | CEO Andy Jassy said AWS grew 36.7% year over year, its fastest growth in 18 quarters, and that Amazon’s AI and chips businesses each exceeded annualized run rates of $25 billion. | These are management-reported figures. They demonstrate commercial momentum but do not show that Amazon’s stock is undervalued. |
NVIDIA founder and CEO Jensen Huang said in the August 26, 2026 earnings release: “AI has reached its inflection point. It’s doing useful work. Its tokens are productive and profitable. Now, compute is revenue.” That is an executive’s view of the opportunity, not independent proof that returns will meet investors’ expectations.
How to judge whether a particular AI stock is worth buying
1. Define the question in valuation terms
Replace “It fell, so is it a bargain?” with “What future cash flows am I paying for, and what must go right for the expected return to be adequate?” Compare the current share price with earnings, free cash flow and a range of realistic growth outcomes. Use both a base case and a downside case.
For a fast-growing company, a low trailing price-to-earnings ratio can be misleading if earnings are temporarily elevated or if infrastructure costs have not yet reached the income statement. A price-to-sales ratio can also mislead when each dollar of sales requires unusually heavy capital spending. Enterprise-value-to-operating-cash-flow, free-cash-flow yield and projected earnings growth can provide additional context, but no single multiple resolves the analysis.
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Current comparable multiples were not established in the available material, so do not substitute a stale ratio or a number copied from an older article. Record the date, share price, fiscal-year estimates and assumptions for every metric you use.
2. Separate demand from paid, profitable usage
High interest in AI demonstrations is not the same as durable revenue. Look for evidence that customers are paying, renewing and expanding usage. For cloud providers, examine booked capacity, usage growth and the relationship between revenue and operating costs. For infrastructure suppliers, examine order visibility, product mix, lead times and whether customers are building capacity faster than they can monetize it.
Reported revenue growth is an important starting point, but ask whether growth comes from a broad customer base or a few very large buyers, whether it is recurring or project-based, and whether the company can maintain pricing as competitors add capacity.
3. Track gross margin, operating margin and cash conversion together
AI services can increase revenue while reducing margins if inference, networking, electricity, depreciation and staffing costs rise faster than prices. Microsoft provides a concrete example: its Microsoft Cloud gross margin percentage was 66% in FY2026 Q3 and decreased amid continued AI-infrastructure investment and growing AI usage, with efficiency gains only partly offsetting the pressure.
Review operating cash flow and free cash flow, not just adjusted earnings. Reconcile stock-based compensation, working-capital swings and capitalized infrastructure. A business that reports strong earnings but repeatedly consumes cash to maintain capacity may require a higher discount rate or a lower valuation multiple.
4. Measure capital intensity and the expected return on new capacity
AI infrastructure requires data centers, power, chips, networking, cooling and specialized equipment. Capital expenditure can remain high for years before new capacity produces its target return. Examine planned spending, depreciation, lease commitments, construction timelines and utilization assumptions.
Microsoft said calendar-year 2026 capital expenditure expectations were approximately $175 billion after a change in lease mix and that FY2027 capital expenditure was expected to grow year over year. Treat those statements as management guidance from its FY2026 Q4 call. The relevant investor question is not simply whether spending is large; it is whether incremental spending should generate returns above the company’s cost of capital within a reasonable period.
5. Test concentration, supply and execution risk
Infrastructure businesses can be constrained by factors outside management’s control. NVIDIA’s regulatory filing identifies land, power, data-center shells and capital availability as important to customer and partner buildout, warning that shortages could affect future revenue and financial performance. A strong order book cannot be converted into sales if customers cannot obtain power or complete facilities.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Also assess dependence on a small number of customers, suppliers, geographic markets or product generations. A delayed platform launch, a competing accelerator, export restrictions or a customer’s decision to design in-house hardware can change the earnings path quickly.
6. Match the stock to your horizon and risk tolerance
A long-term investor may accept several years of volatile investment spending if the expected competitive advantage is durable. Someone needing the money within a few years may not be able to tolerate a prolonged valuation reset, even if the business eventually succeeds.
Set a maximum position size before buying. Decide what evidence would invalidate your thesis—such as slowing paid usage, sustained margin deterioration, cancelled capacity or a balance-sheet strain—and what evidence would merely represent normal volatility. This prevents a falling share price from becoming the only reason to add.
Company-by-company considerations
NVIDIA: exceptional growth with infrastructure and cycle risk
NVIDIA’s fiscal Q2 2027 results show the scale of current demand: $96.2 billion of quarterly revenue, up 106% year over year, including $89.0 billion of Data Center revenue, up 117%. Those are unusually strong reported growth rates, and they explain why investors view the company as a central supplier to AI buildouts.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Repair Windows errors before they cause bigger problems3Fix the driver behind crashes, sound loss and screen glitchesThe investment case still depends on future growth, product transitions and customers’ ability to deploy the systems they have ordered. NVIDIA’s SEC filing for the quarter ended July 26, 2026 highlights the need for land, power, data-center shells and capital. If shortages delay customer projects, revenue can shift between quarters or fall short of expectations. If customers overbuild or develop alternatives, demand and pricing could weaken after a period of rapid expansion.
For NVIDIA, analyze whether the current price already assumes years of exceptional growth, how quickly new architectures can replace older products, the durability of gross margins and the concentration of sales among major data-center customers. The reported quarter is evidence of operating momentum, not evidence of a margin of safety.
Microsoft: diversified cash generation, but AI delivery costs matter
Microsoft’s FY2026 Q3 revenue grew 18% year over year, giving investors a broader earnings base than a pure-play hardware supplier. Its cloud platform, productivity products and enterprise relationships can distribute AI features across an existing customer base.
That diversification does not eliminate execution risk. Microsoft Cloud’s 66% gross margin percentage decreased as the company invested in AI infrastructure and supported rising AI usage. The result illustrates why a successful product launch can initially increase costs faster than revenue. Investors should monitor whether paid AI features, cloud consumption and efficiency improvements eventually expand incremental margins.
Free tools Windows power users keep installed
One-click scans. No signup required.
Management’s FY2026 Q4 call projected approximately $175 billion of calendar-year 2026 capital expenditure after a lease-mix change and said FY2027 capital expenditure would grow year over year. It also described continued double-digit revenue and operating-income growth expectations for FY2027. These are forward statements, so compare later reported spending and margins with the original outlook rather than assuming it will be achieved.
Best Value
Amazon: AWS acceleration inside a multi-business company
Amazon CEO Andy Jassy said AWS grew 36.7% year over year in Q2 2026, the fastest growth in 18 quarters. He also said Amazon’s AI and chips businesses each exceeded annualized run rates of $25 billion. Those figures suggest that AI-related demand is becoming commercially meaningful within AWS and Amazon’s broader operations.
Because Amazon includes retail, advertising and other activities, evaluate AWS economics separately where disclosures allow. Examine operating-income contribution, infrastructure spending, utilization and pricing. An AI revenue run rate is not the same as annual profit or free cash flow, and a rapidly expanding service can require substantial upfront investment.
The central question is whether AWS growth and the AI and chips businesses can produce returns that justify the capital committed, while the rest of Amazon supports or distracts from that objective. Do not conclude that the shares are cheap from management’s growth figures alone.
The spending question: can the industry earn back the investment?
An Associated Press report dated June 23, 2026 said Alphabet, Amazon, Meta and Microsoft planned up to $720 billion of 2026 spending, primarily on AI data centers. The figure describes plans reported at that date, not an audited total of spending already completed. It captures why investors have become more demanding: enormous outlays require enormous and durable profits to earn an attractive return.
High spending is not automatically destructive. Building scarce capacity can create a competitive advantage if demand remains strong and utilization rises. The danger is a mismatch between the useful life of equipment and the speed of technological change, or between capacity commitments and customers’ willingness to pay. Compare each company’s spending with contracted demand, utilization, depreciation schedules and cash generation rather than treating the industry total as either proof of a bubble or proof of inevitable success.
A practical pullback-buying process
- Write the thesis before looking at the chart. State the business role, the source of AI revenue, the expected growth period, the margin assumption and the risks that could disprove the thesis.
- Update the market data. Record the current share price, market capitalization, enterprise value, trailing and forward earnings, free-cash-flow yield and the actual percentage decline from the relevant high. Date every figure.
- Read the latest primary filings and earnings release. Check revenue by segment, customer concentration, gross and operating margins, cash flow, capital expenditure, leases, debt and management guidance.
- Build three scenarios. In a base case, use achievable growth and margin assumptions. In an upside case, specify what additional paid usage or capacity utilization is required. In a downside case, model slower demand, lower prices, delayed capacity and higher costs.
- Compare expected return with risk. Ask whether the potential return compensates you for valuation compression, execution failure and the possibility that capital remains tied up for years.
- Choose an entry method. If the thesis is strong but near-term uncertainty is high, divide the intended allocation into predetermined tranches instead of treating one day’s price as a precise bottom. If the valuation does not provide a margin of safety, waiting is a valid decision.
- Set review triggers. Revisit the thesis after earnings, major guidance changes, material product delays, customer losses or evidence that AI usage is not converting into profitable cash flow. Do not add solely because the price fell further.
Risks that can make a pullback worse
- Valuation compression: Higher interest rates or lower growth expectations can reduce a stock’s multiple even when revenue continues to rise.
- Capacity bottlenecks: Power, land, data-center shells, networking and financing can delay deployments and shift revenue.
- Margin pressure: Compute, energy, depreciation and support costs can grow faster than AI prices.
- Customer concentration: A few large buyers may account for a material share of demand and possess negotiating power.
- Technology substitution: In-house chips, competing accelerators or more efficient models can reduce demand for a particular product.
- Capital-allocation risk: Management may spend aggressively before returns are demonstrated, increasing lease obligations or debt.
- Expectation risk: A company can deliver excellent results and still fall if investors expected even better results.
What to update before making a purchase
Before acting, gather the current share price and drawdown, the latest quarterly and annual filings, updated earnings estimates, free-cash-flow figures, capital-expenditure guidance and segment disclosures. Confirm whether any number is historical, a one-time item, a run rate, a management forecast or an audited result.
Then check whether the valuation still works if growth slows, margins remain below the long-term target and infrastructure spending stays elevated. If the thesis depends on one optimistic assumption—such as uninterrupted triple-digit growth, unlimited power availability or immediate AI monetization—the position is speculation rather than a robust investment case.
PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteBottom line for investors
AI stocks can still be worth buying after a pullback, but only selectively. NVIDIA’s reported growth, Microsoft’s cloud scale and Amazon’s AWS acceleration show real commercial demand, while their disclosures also show supply constraints, lower cloud margins and enormous capital requirements. The right response to a decline is disciplined re-underwriting: identify the business model, test paid demand and cash generation, measure the return on new infrastructure, and compare the updated valuation with conservative scenarios.
Without current prices and comparable valuation data, no responsible analysis can label a specific AI stock a bargain. A pullback is an invitation to do that work—not a substitute for it.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
