The strongest fact: the memory buildout is visible in the filings
Much of this period's sell-side coverage clusters around AI infrastructure: memory and chips, data centers, dispatchable power and quantum computing.1 Most of that coverage is opinion. Micron's numbers are different, because they come from SEC filings and have been checked against that source.
Micron's capital expenditure was $7.676 billion in fiscal 2023 and $8.386 billion in fiscal 2024. It then jumped to $15.857 billion in fiscal 2025.2 That is roughly 1.9 times the prior year's spending, after a year in which spending barely moved. The quarterly figures show the same acceleration. First-quarter capex was $1.796 billion in the 2024 quarter, $3.206 billion in 2025 and $5.389 billion in 2026.2 The 2026 quarter is three times the 2024 one.
Micron's cash balance tells the other half of the story. Third-quarter cash was $7.594 billion in 2024 and $10.163 billion in 2025.3 By the third quarter of 2026 it was $24.995 billion.3 That is about 3.3 times the 2024 level. Between the second quarter of 2025 ($7.552 billion) and the second quarter of 2026 ($13.908 billion), cash had already nearly doubled.3
Cost of revenue, the direct cost of producing what Micron sells, rose more gently. It went from $16.956 billion in fiscal 2023 to $19.498 billion in fiscal 2024 and $22.505 billion in fiscal 2025.4 That is about a third higher over two years, while capital spending roughly doubled. The data we hold does not include Micron's revenue, so we cannot say what happened to margins, and we will not guess. What the filings do show is a company spending far more on capacity while also holding far more cash.
This matters to the AI field because Micron is a supplier at the center of the compute stack. Our entity graph records it as supplying Amazon Web Services and Meta Platforms, and as located in Taiwan.5 If you hold a broad US stock index fund you may own Micron indirectly, but our data does not say how much, so we cannot tell you your exposure.
Where the enthusiasm gets ahead of the evidence
The checked numbers support the idea that the physical buildout is real. They say nothing about whether every company riding the theme deserves its price. Several do not look cheap on the analysts' own words.
The clearest case is quantum computing. Analyst Melissa Tucker puts D-Wave Quantum (QBTS) at a stretched 171 times its sales, with projected 2026 revenue near $35 million. She sets a fair value of $9.50 a share, which she says implies roughly 40% downside, and cites a delayed technology roadmap. She also points to D-Wave's reliance on SkyWare while IonQ's acquisition of that supplier is pending.6 Our entity graph lists IonQ as a direct competitor of D-Wave, so one rival is acquiring a supplier D-Wave depends on.5
The market has not settled on an answer. D-Wave shares fell about 10% on Tuesday, July 28, to $17.64, a drop of $1.88, after a 20% surge the previous session. That surge followed an expanded AT&T partnership and a bullish analyst initiation. The stock was still down 25% over the past month.6 Note also that $17.64 sits well above Tucker's $9.50, so the gap between price and her fair value is wider than her 'roughly 40%' wording suggests. We have not been able to confirm which reference price she used.
Power is the other bottleneck for AI data centers, and the coverage here is bullish with caveats. Analyst Nicholas Amicucci argues that Bloom Energy's ability to provide reliable, dispatchable power to a volatile demand profile differentiates it from competitors. The same piece highlights Bloom's inclusion in the Russell 1000 index and an expansion of its AI infrastructure partners.7 The source is a column called 'Bull of the Day', so treat it as advocacy. Our reliability measure for that source shows 46% of 520 checked claims held up.7
For NuScale (SMR), the dossier records that the shares have fallen nearly 40% since 2026. Bank of America analyst Rinny Singh reiterated a buy rating in early August with a $12 target, implying roughly 24% upside over 12 months.8 The article carrying this call comes from a source where only 22% of 18 checked claims held up, a small sample but a low score.8 We would weigh the analyst's call more heavily than the headline built on it.
The narrative also flags digital-infrastructure strength elsewhere. Digital Realty raised its FFO guidance (funds from operations, the profit measure data-center landlords report), and SpaceX has fallen more than 40% from its highs.1 We hold no source document for either, so these rest on our topic synthesis rather than a checked filing. Our entity graph places SpaceX as a competitor of both OpenAI and Anthropic.5
On chips, ASML stock rose over 4% on September 4 in a sympathy rally. A highly bullish analysis of another top chip name set it off, not any news about ASML itself.9 TD Cowen covers ASML in our graph.5 A move like that tells you about mood, not about ASML's business.
The other side of the economy
The AI story sits alongside a consumer picture that analysts describe as choppy. Bank of America's Lorraine Hutchinson lowered Nike's price target because its sales recovery is taking longer than expected.10 Levi Strauss drew a more positive read. Kendall Toscano reiterated a Buy and a $27 target, citing a brand moving toward 'higher-quality, consistent growth'.11 Capital One's Stephen Biggar notes that its reserve build may have been read negatively because it signaled 'that they expect some weakness or deterioration in credit quality'.12 The same article says Capital One has put more than $35 billion into acquiring Discover and has racked up $1.8 billion in integration expenses.12 Wendy's fell nearly 6% on September 16 after Seaport Global's Eric Gonzalez initiated coverage at neutral, extending a decline since a take-private plan collapsed.13
Why does this belong in an AI story? Because the AI buildout is being financed and valued in the same market as these companies. If households and credit weaken, the stretched valuations flagged above have less room for error. That is our reading, not a claim any analyst here makes.
Event risk on the calendar
Biotech adds its own dates. The FDA has set a target decision date of November 14, 2026 for Summit Therapeutics' ivonescimab application.14 Goldman Sachs analyst Andrea Newkirk resumed coverage of Iovance Biotherapeutics with a buy and a $15 target, and the shares rose.15 Separately, $500 million of notes at 2.9% mature in March 2027 and are expected to be refinanced at higher rates.16 The dossier does not name the issuer.
A note on how far to trust the sources
The articles behind many of these calls come from a news feed where roughly 57% of checked claims held up, whether the sample was 881 or 4,956 claims.6,10,11 That means about four in ten could not be confirmed. The Micron figures are the exception, because they come straight from SEC filings.
What to watch
- Whether Micron's capex keeps climbing from $5.389 billion in a single quarter, and what happens to cost of revenue as it does.2,4
- Whether IonQ's acquisition of D-Wave's supplier closes, and how that affects the roadmap risk Tucker flags.6
- The FDA's November 14, 2026 ivonescimab decision.14
- The March 2027 refinancing at higher rates.16