RUM Group, the Nasdaq-listed company formerly called Rumble, said on 11 November that its Quake AI compute business brought in $68.2 million in the third quarter, more than 70% of total revenue of $96.6 million. In the same quarter, amortization and depreciation across the company reached $85.2 million, up from $3.9 million a year earlier.
That works out to about $3,100 of Quake AI revenue for each of the approximately 22,000 NVIDIA GPUs the company says it runs, or about $1.65 per GPU-hour in use at the approximately 85% utilization it reported, by our arithmetic.
Business Pill · Capex and Depreciation Explained
A short explainer on capital spending and depreciation: how money spent on machines shows up as a cost over the following years. It teaches the general idea only and does not discuss RUM Group.
The key insight: As we read the release, Quake AI’s compute services, from GPU cloud to hosting, became most of RUM Group’s business in one quarter: $68.2 million, more than 70% of revenue. Spread over about 22,000 GPUs, that is about $3,100 per GPU by our arithmetic, an average that includes non-GPU services.
In its third-quarter release, RUM Group reports revenue of $96.6 million, up 290% from a year earlier and 139% from the second quarter. It calls the period “The first profitable quarter in the Company’s history on an Adjusted EBITDA basis”.
On a GAAP basis the release shows a loss from operations of $84.6 million and a net loss of $89.9 million, of which $79.9 million is attributable to RUM Group. Quake AI was in its first full quarter, the release says.
Chief executive Chris Pavlovski said “Quake AI is now the engine of our business” and “Our 22,000-GPU estate is nearly sold out at 85% utilization”.
RUM Group third-quarter figures, 2025 and 2026, from its 11 November results release (unaudited). Amortization and depreciation is reported for the whole company, not by segment.
The GPU Arithmetic
The quarter ran 92 days, or 2,208 hours. Spread over approximately 22,000 GPUs, $68.2 million is about $3,100 per GPU, or about $1.40 per installed GPU-hour; at approximately 85% utilization it is about $1.65 per GPU-hour in use, by our arithmetic. Quake AI also sells CPU compute, storage and networking, so these are revenue averages, not prices.
The release attributes $30.0 million of the rise in cost of services to Quake AI data center costs. Quake AI revenue less those costs is $38.2 million, before depreciation, by our arithmetic.
Amortization and depreciation, reported for the whole company, was $85.2 million, which is $17.0 million more than Quake AI’s revenue, by our arithmetic. Property and equipment stood at $899.7 million on 30 September, against $16.2 million at the end of 2025, the balance sheet shows.
Guidance and Capacity
RUM Group guides fourth-quarter revenue to $88 million to $94 million. The midpoint, $91 million, is about 6% below the third quarter’s $96.6 million, by our arithmetic.
The release also says “Due to market conditions, a certain portion of our previously disclosed 250MW of unmonetized capacity targeted for 2027 may be pushed out beyond 2027”. It reports total liquidity of $142.3 million and convertible notes payable of $358.8 million.
The company raised its stake in Northern Data from 85.2% to approximately 98% and has started squeeze-out proceedings under German law, expecting 100% ownership in early 2027, according to the release.
The Structural Read
As we read the statements, the quarter’s cost line is the hardware: amortization and depreciation of $85.2 million, against $3.9 million a year earlier, is larger than Quake AI’s $68.2 million of revenue, though the release does not split it by segment.
As we read the guidance, $88 million to $94 million for the fourth quarter, with part of the 250MW targeted for 2027 possibly pushed out, means the release itself does not describe further revenuegrowth in the next quarter.
Chris Pavlovski, CEO of RUM Group, 11 November 2026
“We have real assets running real AI workloads for real customers and generating real cash”
Three Implications
FOR GPU OWNERS At approximately 85% utilization, $68.2 million is about $1.65 per GPU-hour in use, by our arithmetic.
FOR THE BALANCE SHEET Property and equipment rose to $899.7 million from $16.2 million at the end of 2025, the release shows.
FOR RUMBLE VIDEO Revenue outside Quake AI was $28.4 million, by our arithmetic, against $24.8 million of total revenue a year earlier.
The framework records: “Companies that solve the power equation will dominate the AI era; those that don’t will watch their chips sit idle”.
As we read it, the framework ties value to chips that are kept busy; RUM Group reports approximately 85% utilization for its GPUs in the quarter and says part of its 2027 capacity may be pushed out because of market conditions. The framework predates the release and does not discuss RUM Group.
What Is Not Established
The release does not split amortization and depreciation by segment, so how much of the $85.2 million sits with Quake AI is not stated. It does not name Quake AI’s customers or say how much of the GPU revenue comes from Rumble, which it calls the original tenant of Quake AI. The per-GPU figures are our averages on approximate counts and include non-GPU services. How much of the 250MW moves beyond 2027 is not quantified.
This Week In The AI Supercycle
The Great Convergence of Capital, Power, and Geopolitics
This week’s issue: for two years the AI build-out was priced as a supply story, and last week as a credit story. This week it was priced as a permission story, and the answer arrived as a cost, not a ban.
RUM Group’s Quake AI produced $68.2 million in its first full quarter, about $3,100 per GPU by our arithmetic, while company-wide amortization and depreciation reached $85.2 million and the GAAP net loss was $89.9 million. The company guides the fourth quarter to $88 million to $94 million and says part of its 2027 capacity may slip.
A note on sourcing. We read RUM Group’s third-quarter 2026 results release in full on 11 November 2026, including its unaudited statements of operations, balance sheet and cash flows. Per-GPU, per-hour and margin figures are our arithmetic on the release’s approximate GPU count and utilization; they are averages, not prices. We did not listen to the conference call, and RUM Group was not contacted.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
