How to Analyze a Technology Stock

The sector label is a trap

The first thing to do with any tech company is not to open a spreadsheet. It is to identify which of five fundamentally different businesses you are looking at. The word “technology” spans companies with gross margins of 25% and companies with gross margins of 85%. It spans businesses that earn recurring subscription revenue and businesses that ship physical chips into a demand cycle that reverses every three years. Treating them under the same analytical framework does not produce a more complete picture. It produces the wrong answer. A metric that signals undervaluation at one type of company signals structural distress at another.

The mistake retail investors make most consistently is importing a valuation framework from a familiar name and applying it to something structurally different. Nvidia trades on book-to-bill ratios and wafer-start scheduling. Salesforce trades on net revenue retention and contract expansion within the existing customer base. Apple trades on product cycle timing, services attach rates, and installed-base monetization. Each of these is a different analytical project, starting from different inputs and watched with different leading indicators.

Before touching a balance sheet or a multiple, place the company in one of five categories: SaaS and cloud, semiconductors, hardware and devices, platforms and marketplaces, or AI infrastructure. Each has a different cost structure, a different revenue pattern, and a different set of metrics that actually predict where results are heading. The rest of this page runs through each one.

The five tech business models

SaaS and cloud companies sell software as a subscription. Revenue arrives monthly or annually, recognized ratably over the contract term. The cost to service an additional customer is very low once the software is built, which is why gross margins at pure-play SaaS businesses typically run 70 to 85 percent. The critical question at a SaaS company is not whether it is growing, but whether customers are staying and expanding. Net revenue retention (NRR) above 110 percent means existing customers are spending more year over year, creating organic growth before a single new logo signs. An NRR below 100 percent means the existing base is shrinking — and no amount of new logo adds fully compensates for a leaky bucket. Salesforce ran NRR around 120 percent at its peak growth phase: 20 percent organic revenue growth from customers already on the platform, before new sales touched it.

Semiconductor companies design or manufacture chips. The business is capital-intensive: TSMC spent roughly $36 billion on capex in 2023, more than the annual revenue of most mid-cap software businesses. Revenue cycles are driven not by end demand in the simple sense, but by inventory cycles — customers overbuy during scarcity, then draw down stockpiles, then stop ordering even as underlying demand holds steady. The leading indicator is book-to-bill ratio, the ratio of new orders to billings shipped. When book-to-bill drops below 1.0, the order pipeline is contracting faster than current shipments. That is typically the earliest visible warning that a revenue air pocket is six to nine months out, before it appears in reported results.

Hardware and devices companies sell physical products, usually on a one-time purchase basis. Revenue is episodic, tied to product release cycles. The economic engine is the installed base: the device is the foothold, and recurring revenue comes from services, accessories, or consumables attached to it. The metric that matters most is attach rate — how much additional revenue the company extracts from each unit sold over its lifetime. A hardware company with a declining attach rate is losing the recurring-economics battle even as unit volumes hold flat.

Platforms and marketplaces are toll-road businesses. They do not own inventory or produce content — they connect buyers and sellers, creators and audiences, and collect a percentage of every transaction (take rate) or a fixed fee for access. Network effects are the structural moat: more sellers attract more buyers, which attracts more sellers. The metric that matters is gross merchandise volume (GMV) growth relative to take rate movement. If GMV is growing but take rate is falling, the platform is subsidizing volume. If both are growing, the flywheel is working. Amazon’s third-party marketplace and Airbnb both run this model; so does Apple’s App Store, which is why the Services segment generates margins above 70 percent on software distributed through a hardware-built network.

AI infrastructure is the newest category and the least settled analytically. It includes data center operators, power and cooling providers, GPU manufacturers, and the hyperscalers building training clusters. The risk unique to this category is the transition from training to inference. Training workloads require the most powerful chips regardless of price. Inference workloads can often run on cheaper, more efficient hardware as models compress and specialize. A company whose revenue depends on training-cluster capex buildouts is in a structurally different position than one whose revenue scales with inference deployment. Most retail investors have not yet learned to separate these two demand pools — and Nvidia’s 2024 revenue acceleration was almost entirely a training phenomenon, a fact that matters enormously for what happens next.

What the income statement hides

Three accounting treatments in tech distort GAAP earnings enough to make healthy companies look impaired and struggling ones look stable. Understanding all three before reading any tech income statement is not optional.

Stock-based compensation is the most commonly misread. The GAAP income statement counts it as an operating expense. Wall Street “adjusts” it out entirely and presents non-GAAP earnings that exclude SBC. Neither treatment fully captures the economic reality. SBC is not a cash outflow in the quarter it is reported — the company does not write a check to a bank. But it is real dilution: the company issues shares to employees, which reduces each existing shareholder’s ownership percentage. Those shares had value at the price they were granted. That value came from somewhere.

The right check is diluted share count growth year over year alongside free cash flow per share. If shares outstanding are growing 9% annually and management is citing strong free cash flow growth, the per-share free cash flow is growing substantially more slowly than the headline number implies. Companies with SBC above 10% of revenue and minimal buyback programs are transferring value away from public shareholders every quarter, in plain sight, under a label that reads “non-cash.”

R&D expensing creates the opposite distortion. Under GAAP, most software development costs are expensed immediately — they hit the income statement in the quarter they are incurred, even if the software being built will generate revenue for five or ten years. A company investing heavily in new product development shows depressed earnings relative to one in harvest mode spending minimally on R&D. This does not mean the high-R&D company is economically weaker — it may be building substantially more durable future cash flows. The practical test: compare R&D as a percentage of revenue against the revenue growth rate. A company spending 30% of revenue on R&D and growing at 25% annually is making a different bet than one spending 30% on R&D and growing at 6%.

Deferred revenue is the distortion that makes cash-generative subscription businesses look worse than they are on GAAP financials. When a software company collects an annual subscription upfront — $120,000 from a new enterprise customer — it books $10,000 per month as revenue and records the remainder as deferred revenue on the balance sheet, a liability. The cash is already in the bank. The income statement has not yet recognized it. GAAP revenue at a growing subscription business consistently understates the cash the business is actually collecting. To see the real cash picture, look at the change in deferred revenue — or go directly to the cash flow statement, which reconciles the gap. Two SaaS companies growing at 20% by GAAP revenue can look identical while one is collecting cash 35% faster than the other, because its customers are signing multi-year deals upfront. In a downturn, that prepaid cash is a cushion the income statement never showed you.

Growth versus profitability: where the tradeoff actually sits

The Rule of 40 is the simplest useful heuristic in software analysis. A healthy software business should have its revenue growth rate plus its operating profit margin add up to at least 40. A company growing at 50% with a -10% operating margin scores 40 — passing. A company growing at 20% with a 25% margin also scores 40. The point is that growth and profitability are substitutable in the short term: a company investing aggressively in growth can operate at a loss and still be creating value, provided the growth is real and the unit economics underneath it are intact.

The Rule of 40 only works if gross margin is structurally sound. That is the number to start with before anything else. Gross margin is the percentage of revenue remaining after direct costs — for software companies, primarily hosting, customer support, and implementation services — and it sets a ceiling on everything downstream. A company with 45% gross margins has an operating margin ceiling well below 45%, because sales, marketing, R&D, and G&A still sit above the operating income line. A company with 80% gross margins can invest aggressively in growth and still reach operating profitability as revenue scales, because each incremental dollar of revenue keeps 80 cents before any growth spending. Gross margin at software companies does not change very much as the business grows; sustained compression is almost always a signal of pricing pressure, growing professional services that carry lower margins, or infrastructure costs outpacing revenue.

The harder question is whether a company burning cash is investing in something real or is burning cash because the unit economics are structurally broken. The test is customer acquisition cost (CAC) versus lifetime value (LTV). If the company spends $50,000 to acquire a customer who generates $180,000 in gross profit over their expected lifetime, the cash burning today is a rational investment in future cash flows. If it spends $80,000 to acquire a customer generating $60,000 in lifetime gross profit, the unit economics are underwater. No amount of scale fixes negative unit economics — it just means losing more money on more customers, faster.

Consider two cloud companies both growing at 30% annually. Company A runs 78% gross margins and -8% operating margins, with NRR of 118%. Company B runs 52% gross margins and -22% operating margins, with NRR of 95%. Company A’s Rule of 40 score is 22 (30 minus 8) — mediocre on the headline, but the NRR tells you customers are expanding, the gross margin is structurally excellent, and the operating losses are a function of investment in sales capacity, not of broken unit economics. Company B’s Rule of 40 score is 8. More critically, 52% gross margins in software suggest either a heavy professional services load, significant third-party licensing costs, or infrastructure that is not scaling. The NRR below 100% means the existing base is contracting. Both companies appear identical on headline revenue growth. The analytical picture is not close.

The metric that fits your company type

NRR, gross margin, Rule of 40, book-to-bill, deferred revenue trends — these do not all apply to every tech company. NRR is the most important single number at a SaaS business and produces no useful signal at Qualcomm. Book-to-bill matters enormously at a semiconductor equipment company and is irrelevant at a cloud platform. Applying the longest possible checklist to every tech company is not more rigorous. It is more confusing, and it buries the signals that actually matter under ones that do not.

The right sequence is taxonomy first: identify the business model, then apply the metrics that fit that model. The tool below runs through the six most important metrics for each of the five sub-sectors, with benchmarks and red flags attached. Click through all five categories. The point is not to memorize the lists. It is to build the intuition for why the same revenue growth rate means something completely different at a chip company than at a subscription software company — and why confusing the two is the mistake that most often precedes a painful lesson.

Red flags that show up before the guidance cut

The stock price usually drops before management acknowledges a problem. But the signals appear in the filings first, for investors willing to look at the right numbers. These are the patterns that consistently precede a tech company cutting guidance or missing a quarter by a wide margin.

• Rising days sales outstanding (DSO). DSO measures how long it takes a company to collect on its invoices. When DSO rises significantly — from 55 days to 75 days over two quarters — it can mean customers are under financial pressure, or that the company is booking revenue from deals not fully closed, a pattern that precedes restatements more often than management commentary acknowledges. Normal quarterly fluctuation is 10 to 15 percent. Sustained upward movement is the flag.
• Customer concentration creeping above 20%. If a single customer represents more than 20% of revenue and that customer is renegotiating, shifting to a competitor, or building capability in-house, the damage is large and fast when it arrives. This is disclosed in 10-K filings under risk factors and customer concentration sections. Most retail investors read it after the event.
• Gross margin compression blamed on “mix shift.” Mix shift is real — if a software company sells more professional services than expected, blended gross margin falls because services carry lower margins than software. But sustained compression over multiple quarters, with the explanation consistently being mix, is usually a sign that the software itself is facing pricing pressure, not that services happened to be in unexpected demand. Watch for the explanation staying constant while the margin direction stays negative.
• R&D headcount cuts. Tech companies protect engineering talent above almost everything else, because that capability takes years to rebuild once lost. When a company reduces R&D headcount — visible in total headcount disclosures, layoff announcements, and LinkedIn data — it is signaling that near-term cash pressure is more severe than guidance language is communicating. The long-term product pipeline is being traded for short-term margin support.
• Deferred revenue growth decelerating sharply. At a subscription software company, deferred revenue is a forward indicator of future recognized revenue. If deferred revenue growth slows materially before revenue growth slows — which it will, because revenue recognition lags billing — the reported revenue numbers are about to follow. A deferred revenue deceleration today is often a revenue miss in two to three quarters. This is visible in the balance sheet every single quarter, not just annually.

None of these signals guarantees a bad quarter. They narrow the probability distribution. A company showing three of the five simultaneously is not in a comfortable position, regardless of what the guidance range says. The guidance range is management’s view of the future. The signals in the filings are the present tense.