Rumble Becomes RUM Group: From Speech Platform to AI Compute Aspirant
Northern Data closes, Quake AI is born, and management issues first ever revenue guidance — betting 250MW of power and a video corpus on the agentic AI era
RUM · Earnings Call · 2026-08-10
From Speech Rails to AI Compute Racks
The Rumble that reported on August 10 is no longer simply the video platform that battled Big Tech over censorship. Rumble Inc. renamed itself RUM Group and closed its acquisition of Northern Data on June 17, pivoting decisively into the AI infrastructure business. Chris Pavlovski's framing was unambiguous:
Last quarter, I told you this would be the last call before Rumble meaningfully entered the cloud and agentic AI era. Today, I'm glad to say that transformation is complete.
The transformation isn't cosmetic. RUM Group now operates two units — the Rumble video platform and Quake AI, combining Rumble Cloud with Northern Data's estate of roughly 22,000 NVIDIA H100/H200 GPUs. Management is leaning hard into AI compute as the financial engine: the GPU estate is running above 85% utilization, and in June the company signed a multiyear agreement with Together AI to deploy NVIDIA HGX B300 capacity — validation of execution credibility and a signal NVIDIA is willing to allocate supply.
The biggest number on the table is 250 megawatts of currently unmonetized power targeted for 2027, anchored by a 180MW site near Atlanta, which management frames as a $3B+ annual run-rate opportunity. In the Q&A, Mike Masci defended AI-compute-as-a-service versus the "powered shell" model (leasing shells at ~$1.5-2M per megawatt-year versus ~$11M per megawatt-year delivering AI compute). This is a company riding a broad wave visible across the global tape — hyperscale datacenter build-out, GPU scarcity — but from a very different starting point.
The Video Data Frontier: A Second Derivative
If compute is the first act, the more company-unique idea is what the video corpus becomes. Pavlovski's thesis: as AI moves into the robotic era, spatiotemporal data — i.e., video — will be the next scarce fuel, and Rumble's trove of it, combined with Quake AI's rails, gives RUM a position "unlike today's neoclouds": “As the AI industry moves into the robotic era in the coming years, Rumble's spatiotemporal data, otherwise known as video data, becomes increasingly valuable and very important to robotic learning.” — Christopher Pavlovski, Founder, Chairman and CEO · 2026-08-10 He pointed to Reddit's data-licensing value as precedent, and noted Quake AI's own clients have already expressed interest in Rumble's video data. This theme — "video data," "spatiotemporal data," "robotic era" — spikes for the first time in Q2 2026.
The contrast with the prior year's narrative is stark: earlier calls were about the Tether ad commitment, Rumble Wallet, and creator monetization. Now the headline metric set itself is being retired. Masci put it plainly:
“This will be the last quarter we present MAU and ARPU as headline metrics for the company.” — Michael Masci, CFO · 2026-08-10
The company is moving to segment reporting (Rumble Video and Quake AI) and, for the first time, issued formal revenue guidance — Q3 2026 revenue of $87M-$93M, roughly double the record $40.4M Q2 (up 61% YoY). That is a real departure: as recently as May, management was still hedging, saying in response to analyst questions about the S-1 forecasts that “no, those forecasts are not guidance.” — Michael Masci, Chief Financial Officer (CFO) · 2026-05-15 Now the discipline is the point — Masci: "Building a disciplined, credible guidance process is something I'm personally focused on."
The Numbers Behind the Pivot
The scale of the step is worth grounding in the fundamentals. The latest filed 10-Q (May 14) predates the Northern Data close, so the picture is of a pre-transaction Rumble: total revenue was roughly $25M (Q1 2026), with gross margin still slightly negative at -6%. The video business averaged low-to-mid $20Ms per quarter over the past year — which is precisely why the $87-93M guide represents a step-function, not an increment.
The market has already voted: the stock is up 81% over the last 90 days to near its May 2026 peak. But the pivot carries real risk. The quarter's net loss widened to $80.3M from $30.2M, driven by $28.3M of acquisition-related transaction costs and higher depreciation; the company still burns cash. The dependence on one strategic partner remains — Tether contributed roughly $4.8M of Q2 revenue, and management confirmed the ad commitment is live (“The Tether ad commitment has already begun this quarter” — Christopher Pavlovski, Founder, Chairman and CEO · 2026-05-15). And the entire scarcity thesis rests on pricing power persisting well into next year: “I don't see any point in which we're going to have capacity meet that demand in the next 1 to 2 years.” — Christopher Pavlovski, Founder, Chairman and CEO · 2026-08-10
What Changed, and Why It Matters
The simplest read: RUM Group is no longer a video company with a small cloud — it is an AI infrastructure company that also owns a video platform with a unique data asset. The genesis of Quake AI is entirely new in Q2, while agentic AI and unmonetized capacity frame the roadmap. The prior quarter's Q&A centered on the Northern Data close (in March, Pavlovski noted "there is capacity to scale immediately in some of their data centers with the GB300s"); now that the deal is closed, the question is execution — converting megawatts into contracted, revenue-generating compute.
That is the crux for investors. The company has given the market a hook it previously avoided (formal guidance, megawatt targets, a $3B+ run-rate claim), and the tape has responded. Whether Quake AI's utilization and Together AI-type contracts can compound — and whether the video-data licensing thesis ever becomes more than a "bucket of potential revenue" for the creator community — will determine whether this is a genuine inflection or a well-sold story.