Rumble Closes Northern Data Acquisition, Adding 22,000 Nvidia H100/H200 GPUs and 250 MW of Power Capacity

Rumble has closed its acquisition of Northern Data AG, picking up 22,000 Nvidia H100 and H200 GPUs distributed across nine data centers and up to 250 megawatts of power capacity, according to Morningstar/MarketWatch (published 2026-06-18 GMT).

The deal transforms Rumble's infrastructure profile overnight. Prior to the acquisition, the video platform's compute footprint was sized for media delivery — transcoding, CDN, storage — not for the dense, low-latency GPU clusters that AI inference and training workloads demand. Northern Data, a Frankfurt-listed HPC and AI infrastructure operator, brings exactly that: purpose-built GPU capacity at a scale most hyperscaler tenants would recognize.

The hardware numbers warrant scrutiny. Twenty-two thousand H100 and H200 units is a meaningful cluster. Nvidia's H100 SXM5 delivers roughly 3.9 petaFLOPS of FP8 throughput; at that density, Rumble's acquired fleet could theoretically post aggregate throughput in the exaFLOPS range across the full stack, depending on interconnect topology and utilization. Power capacity of 250 MW is similarly substantial — a rough rule of thumb puts a densely loaded H100 rack at 10–20 kW per GPU, implying significant headroom for expansion or third-party co-location and cloud revenue.

That power figure matters as much as the chip count, possibly more. GPU availability has loosened somewhat as Nvidia's supply chain has scaled, but locking in hundreds of megawatts of contracted power across nine sites is a multi-year infrastructure moat. Data center power is a constrained resource across every major market, and pre-permitted, energized capacity doesn't come quickly or cheaply.

The strategic read here is straightforward: Rumble is positioning itself as a vertically integrated AI compute provider, not merely a video-hosting platform with AI ambitions bolted on. Whether the company can monetize that stack — through internal workloads, external cloud services, or both — is the open question. The CEO's public insistence, reported by MarketWatch, that this isn't a fad-driven "pivot to AI" is notable precisely because the market has seen that movie before. Media companies reaching for GPU assets have a mixed track record; the ones that succeed typically have either captive internal demand to justify the capex or a credible path to third-party revenue.

Rumble has one plausible internal demand driver that many of those prior pivots lacked: a large, politically distinct user base generating native content and engagement data that could feed recommendation, moderation, and search workloads. Whether that's sufficient to absorb even a fraction of 22,000 H-series GPUs at commercial utilization rates is a separate question.

The Northern Data deal also resolves an uncertain situation for the German company, which had been navigating financial pressures at the intersection of high capex and a volatile AI infrastructure market. For Rumble, the acquisition route — buying distressed or repositioning European HPC assets — likely offered a faster and potentially cheaper path to scale than procuring equivalent Nvidia allocations on the open market at current contract prices.

What the close of this deal does not settle is execution. Integrating nine data centers across what was a German-headquartered operation into a U.S.-listed video company's operational structure is a non-trivial undertaking. Power contracts, cooling infrastructure, network interconnects, and staff retention all carry integration risk. The GPU count is the headline; the balance sheet and operational cadence over the next four to six quarters will tell the more consequential story.