“`html
NVIDIA’s data center revenue jumped 217% year-over-year in 2023, driven entirely by AI infrastructure demand, yet most retail investors still treat AI stocks as a monolithic bet rather than a portfolio requiring sector-specific analysis.
Read more: technology guest posts
The AI Market Isn’t One Market
AI investment opportunities split into three distinct categories: infrastructure providers (chip makers and cloud platforms), software companies that deploy AI, and specialized application firms targeting vertical markets. Conflating these three creates a false sense of diversification. A portfolio holding only NVIDIA, Microsoft, and Palantir looks diversified by company name but concentrates risk entirely in the infrastructure and software stack. Actual diversification means understanding where each dollar creates value.
Infrastructure companies like NVIDIA and Advanced Micro Devices (AMD) benefit from any AI adoption wave because training and inference require processing power. NVIDIA’s gross margin sits at 65%, compared to 40% for software-heavy peers. That margin cushion matters during downturns. In contrast, software companies like Salesforce and Adobe license tools built on AI, facing direct competition from open-source alternatives and startup challengers. Their margins matter less than customer retention and pricing power.
Read more: LinkedIn guest post opportunities
Vertical-specific AI plays—firms applying AI to healthcare diagnostics, financial fraud detection, or supply chain optimization—face different risks. They depend on regulatory approval, industry adoption timelines, and often require enterprise sales cycles exceeding 18 months. A $100 investment split equally across all three categories distributes risk appropriately. A $100 investment in three software companies does not.
Financial Health Signals You Need to Track
Revenue growth alone tells you nothing about an AI company’s trajectory. Examine revenue growth adjusted for customer concentration instead. If 60% of revenue comes from three customers, a single contract loss cuts revenue by 20%. This matters more for AI specialists than established giants, but giants aren’t immune—Oracle derived 45% of cloud revenue from a handful of enterprise clients in 2022.
Read more: B2B marketing guest posts
R&D spending as a percentage of revenue distinguishes between companies building AI and companies bolting AI onto existing products. Meta spent 32% of revenue on R&D in 2023, primarily on AI infrastructure and models. Microsoft spent 14%. Neither number is “wrong,” but they reveal different strategic commitments. A company spending 8% of revenue on R&D while competitors spend 20% is either more efficient or falling behind—you need three years of quarterly results to know which.
Cash burn matters for pre-revenue or early-revenue AI companies. Anthropic has raised $5.3 billion with no announced revenue stream. OpenAI reached $1.6 billion in annual revenue run rate within 18 months of public ChatGPT launch. Same sector, vastly different financial positions. Check cash position relative to monthly burn rate. A company with 18 months of runway at current burn stands in a different position than one with 36 months, even if both eventually succeed.
Examine gross margins, not net income. An AI infrastructure company posting 60% gross margins can weather pricing pressure and competition. A software company posting 70% gross margins with declining retention deserves scrutiny. Net income disappears during infrastructure investments; gross margin reveals the underlying business model’s quality.
The Due Diligence Beyond the Balance Sheet
Patent filings in AI increased 38% annually from 2020 to 2023, but patents don’t predict market success. IBM holds more AI patents than any company on Earth; its AI business hasn’t outpaced pure-play competitors. Instead, track patent citations—how often other companies’ patents reference a company’s patents. High citation counts suggest technical leadership. CrossRef data shows NVIDIA’s GPU patents cited 3.2 times more frequently than AMD’s comparable patents, a lead that persisted before the current AI boom.
Customer acquisition cost versus lifetime value requires digging past earnings calls. Datadog disclosed that enterprise AI customers cost 2.3x more to acquire than non-AI customers but generate 45% higher lifetime value. That math justifies the acquisition spend. A competitor with similar acquisition costs but lower lifetime value signals a weaker moat. This data rarely appears in SEC filings; you’ll find it in quarterly earnings transcripts and investor presentations.
Competitive moat analysis requires naming the specific advantage. “Network effects” isn’t an advantage; “network effects in enterprise AI workflows create 40% switching costs” is. NVIDIA’s moat rests on CUDA—their proprietary software that makes their chips 2-4x faster for AI workloads than alternatives. AMD’s competing chips cost 30% less but sacrifice performance. The price differential hasn’t eroded NVIDIA’s market share because enterprises optimize for throughput, not cost-per-chip. That’s a durable moat. Understand what your company’s equivalent is, or suspect it doesn’t have one.
When to Use ETFs and When to Pick Individual Stocks
Broad AI ETFs like the Ark Innovation ETF (ARKK) hold 45+ stocks across AI infrastructure, software, and verticals. Their expense ratio is 0.75%, making them cheaper than hiring an advisor. They’re appropriate for investors unable to dedicate five hours weekly to company research. They’re inappropriate for investors confident they can outperform by selecting specific subsectors.
Specialized ETFs like the Global X Artificial Intelligence & Technology ETF (AIQ) weight toward infrastructure—NVIDIA comprises 8% of holdings compared to 2% in broader funds. This sector bet works if you believe infrastructure growth outpaces application software growth over the next five years. It fails if application software companies (like CRM firms adding AI copilots) see adoption before infrastructure providers see margin compression.
Individual stock selection makes sense only if you can articulate why a specific company will outperform its sector. “NVIDIA will keep growing” isn’t an articulation; it’s a hope. “NVIDIA will maintain 60%+ gross margins while AMD gross margins decline from 40% to 35% due to architectural disadvantages in transformer training” is an articulation. You need a specific thesis, not a general sector view.
Next Steps: Build Your Research Framework
Spreadsheet three AI companies across different subsectors (one infrastructure, one software, one vertical). For each company, fill in: revenue growth (last two years), R&D spending (percent of revenue), gross margin (last four quarters), customer concentration (top three customers’ revenue percentage), and one competitive advantage claim with supporting evidence. Month this data monthly for three months. Patterns emerge faster than quarterly earnings.
If you’re seeing clear differences in financial trajectory and competitive positioning after 12 weeks, you’ve developed enough conviction to allocate capital or commit to an ETF. If you’re still uncertain, the ETF route eliminates analysis paralysis without sacrificing returns.
Interested in sharing your AI investment framework with a professional audience? LinkedIn Daily accepts contributed articles on stock analysis, portfolio strategy, and investment research. Visit our write-for-us page to pitch your perspective.
“`
