Friday, October 30, 2026

Nvidia's Cost of Revenue Has Nearly Quadrupled in Two Years While the AI Buildout Spreads Into Lidar, Agents and Networking

SEC-filed Nvidia figures show cost of revenue rising from $16.6B to $62.5B across fiscal 2024–2026, even as cash stayed in a $7–11B band. Around that core, October 2026 deals show the spending spreading to networking, physical-AI data and agentic software. Only some of those claims have been checked.

LM Salvado
LM Salvado

October 17, 2026

Source Trace Score12 source documents12 with a live linkVerifiability: Strong
Nvidia's Cost of Revenue Has Nearly Quadrupled in Two Years While the AI Buildout Spreads Into Lidar, Agents and Networking
Image generated by AI for illustrative purposes. Not actual footage or photography from the reported events.

The number underneath the story

Nvidia's cost of revenue, the direct cost of producing what it sells, was $16.621 billion in fiscal 2024, $32.639 billion in fiscal 2025 and $62.475 billion in fiscal 2026.1 By our arithmetic, that is nearly four times the cost base in two years. These figures come from SEC filings and are fidelity-checked.1 The company is paying far more to make far more.

One caution matters for ordinary investors. Our verified set contains Nvidia's costs, cash and earnings per share, but not its revenue. We therefore cannot calculate a profit margin here, and we won't guess one. Rising costs are consistent with surging demand. On their own they do not show how profitable that demand is.

Cash did not pile up

Many readers assume a company at the centre of the AI boom is hoarding cash. The filed numbers say something more mixed. Year-end cash was $7.28 billion in fiscal 2024, $8.589 billion in fiscal 2025 and $10.605 billion in fiscal 2026.1 That is steady growth, but small beside the cost figures above.

The quarterly path is uneven. Cash was $15.234 billion in Q1 2026, then $11.639 billion in Q2 2026 and $11.486 billion in Q3 2026.1 It stood at $13.237 billion in Q1 2027.1 So the Q1 2027 balance is below the Q1 2026 peak, even though quarterly cost of revenue climbed from $17.394 billion to $20.458 billion over the same pair of periods.1 The dossier does not say where the cash went, whether into buybacks, investment or working capital. We can see the balance move but not the reason.

A caution on earnings per share

Reported earnings per share were $11.93 in fiscal 2024, $2.94 in fiscal 2025 and $4.90 in fiscal 2026.1 Read naively, that looks like profit per share collapsing while costs quadrupled. We would not read it that way. The dossier does not tell us whether these figures are measured on the same share count, and a change in share count, such as a stock split, would make the series incomparable. Until that is checked against the filings, treat the year-to-year EPS line as unreliable for trend-spotting. This is the kind of seam we would rather show than smooth over.

The buildout is widening

Nvidia's costs are one layer of a stack being built at several levels at once. At the OCP Global Summit on October 12, 2026, Cisco introduced silicon, systems and optics for open AI infrastructure. These included the first G300-powered systems shipping to customers and new 1.6T G302 silicon.2 Cisco and NVIDIA also plan an open specification for next-generation 400G-per-lane pluggable optics.2 Cisco's Kevin Wollenweber framed the shift this way: "AI is redefining what it means to scale infrastructure. It's no longer just about individual data centers or clusters, but how we connect compute at massive scale without compromising performance or resilience."2 For investors, the point is that spending is moving beyond the chip itself to the links between chips.

The next layer is data. On October 14, Hesai launched Kosmo, a 3D world data engine that combines the Kosmo One AI spatial camera with the Kosmo World cloud platform. It is aimed at robot training, world-model development and content creation.3 Hesai also announced partnerships with embodied-AI companies Galbot and Dexmal.3 Hesai's Shaoqing Xiang put the business case this way: "Physical AI needs real-world data that is accurate, affordable and available at scale. Meeting all three requirements has been a persistent problem."3 That is a company's own claim about its product, not independent evidence of demand.

The top layer is software. TELUS Digital announced a partnership with Parloa, becoming an implementation partner that brings Parloa's AI agent platform to enterprise clients.4 Zeta introduced AthenaOS, which CEO David A. Steinberg described as an expansion of Zeta's AI infrastructure "beyond marketing".5 A Snipp survey of 1,000 U.S. adults who regularly use AI for search and shopping found that 61.8% had purchased after an AI recommendation.6 Because it surveys only regular AI users, that figure says nothing about shoppers in general.

How deals are being paid for

Capital structure shows up in the smaller deals too. On September 30, Grindr agreed to acquire PurposeMed, the parent of telehealth provider Freddie, for $250 million in cash and stock, with up to $70 million more in possible cash consideration.7 On October 14, SKYX closed its merger with Deako, issuing 25,000,000 shares that are subject to a 21-month lock-up and leak-out. Post-merger, SKYX shareholders own 84.4% of the combined company.8 SKYX also paid Deako's senior lender $2 million at closing and issued an $8.5 million secured promissory note, with $2.25 million due in Q1 2027 and $6.25 million in Q4 2027.9 Neither deal is an AI-infrastructure story. Both show that stock, cash and deferred notes are being mixed to fund growth.

The governance seam

The commercial momentum sits beside unresolved safety questions at OpenAI. Our topic summary also mentions a safety-lead resignation and leadership pledges on monitoring and open-source timelines, but the verified dossier holds no sourced detail on them, so we make no claims about them. What we do have is a quote from OpenAI's chief research officer, Mark Chen: "I do kind of reject the premise that OpenAI is a company with visible impacts in the world and therefore OpenAI is not training safe and aligned models."10 Treat the sourcing with care. The MIT Technology Review items in our set have a measured reliability of 0% of 11 checked claims holding up, and the quote is undated.10

How much to trust the sources

The press-release sources behind the Cisco, Hesai, TELUS, Snipp and Grindr items carry a measured reliability of 56% of 4,960 checked claims holding up.2 That is a coin flip leaning slightly one way. The SEC-derived Nvidia numbers are the firmest ground in this piece, and the announcements are claims by the companies making them.

What to watch

  • Nvidia's next filing: does cost of revenue keep climbing, and does cash follow? We'll want revenue alongside it to judge profitability.
  • Whether EPS is restated onto a consistent share basis, which would resolve the trend question above.
  • Whether Cisco's G300 systems, shipping now, show up in customer orders.
  • The SKYX note payments due in Q1 and Q4 2027.

Source documents

Via News is a conduit. We point to the source documents behind this report — we don't replace them. Trace any claim to its source and decide what to trust. How we source

Source Trace Score12 source documents12 with a live linkVerifiability: Strong
  1. [1]News articleYahoo Finance· October 3, 2026
    Aetna 2027 Medicare plans offer reliable, affordable access to care & member support
  2. [2]News articleYahoo Finance· October 8, 2026
    Aetna 2027 Medicare Plans Offer Reliable, Affordable Access to Care & Member Support
  3. [3]News articleYahoo Finance· October 14, 2026
    Building the Data Infrastructure for Physical AI: Hesai Launches Kosmo 3D World Data Engine
  4. [4]News articleCB Insights
    CEO Interview: Napo Pet Insurance
  5. [5]News articleYahoo Finance· October 12, 2026
    Cisco Advances Scalability and Openness at OCP 2026
  6. [6]News articleYahoo Finance· September 30, 2026
    Grindr to Acquire Freddie, Expanding Access to HIV Prevention for Millions of Users
  7. [7]News articleYahoo Finance· October 6, 2026
    Snipp Survey Finds AI is Reshaping the Shopper Journey, with 61.8% of AI Users Having Purchased After an AI Recommendation
  8. [8]News articleYahoo Finance· October 14, 2026
    TELUS Digital Partners with Parloa to Help Enterprises Build and Manage AI Agents for Stronger Customer Relationships
  9. [9]News articleMIT Technology Review
    The Download: AI roadblocks for humanoids and portable rubber dams
  10. [10]News articleMIT Technology Review
    The Download: OpenAI’s chief research officer explains its hacking response
  11. [11]News articleMotley Fool
    This AI Token Is Up Nearly 1,800% For the Year. Here's Why It's Getting So Much Attention From Crypto Investors.
  12. [12]News articleMIT Technology Review
    “We’re not going to shoot ourselves in the foot” over hack fallout, says OpenAI’s chief research officer

In this story · Knowledge Files

LM Salvado
LM Salvado

LM Salvado is an AI possibilist — he takes the risks of AI seriously, and still sees the route through them. Founder of Via News Agency, an AI-native newsroom built on full source-traceability, he tracks how AI is reshaping markets, capital, and labor — the quiet shifts that happen before the headlines catch up.

What we know · the intelligence behind this page
Live from the substrate
What we're seeing
ESMO 2026 Oncology Data Wave Builds Toward Ivonescimab PDUFA, While Micro-Cap Biotech Exits the Market
Late October 2026 is a dense catalyst window led by ESMO 2026. Daiichi Sankyo's DXd antibody-drug conjugate (ADC) programs (DESTINY-Lung04, DESTINY-Breast11, REJOICE-Ovarian01/02, TUXEDO-3), Guardant's 25 abstracts and the PACE-LUNG ctDNA trial all report there. These feed into the FDA's 14 November 2026 PDUFA date for ivonescimab. Large and mid-cap developers with late-stage data and commercial launches (ZEGFROVY, ANKTIVA, MIUDELLA, lorundrostat, duvakitug) are drawing capital and attention. Spago Nanomedical's board decision to seek a Nasdaq First North delisting (EGM 19 November, application 25 January 2027) shows the capital squeeze on early-stage, single-asset names (Tumorad is Phase I).
Our read on the data ›
Signals we're tracking
EPKINLY Regulatory-Clinical Success Cascade
High probability of expanded label indications, additional combination approvals, and competitive positioning strength in follicular lymphoma market. Predicts positive commercial uptake and potential accelerated review for related indications.
Patterns we're watching ›
Where sources disagree
Ford Motor Company
Both facts give cash for Ford Motor Company at the same point in time (2025-12-31; FY 2025 year-end and Q4 2025 year-end are the same date). Both are labelled USD, but the values are 23,356,000,000 and 23.356, which cannot both be true in USD. The two numbers agree once the second is read as billions (23.356 bn = 23,356,000,000), so the economic content is consistent and the conflict is a unit/scale error in the Q4 2025 record. The period is not the reason for the conflict: a Q4 and an FY figure at the same year-end date should match.
We flag conflicts openly ›
Recently verified
✓ Checked against the original source
4,994
facts traced to their source — and we flag the ones that don't hold up.
101 entities tracked4,994 facts checked against source5,503 source documents archived
Query this data → isubstrate.com