NETINT VPUs turn retired servers into 320-stream encoders
NETINT advocates recommissioning decommissioned servers with its Video Processing Units (VPUs) to expand video encoding capacity. The approach offloads H.264, HEVC, and AV1 encoding to low-power VPUs, achieving up to 32 1080p30 streams per card. This method targets organizations facing GPU/server supply chain delays and seeks to free new servers for AI workloads.
Key Takeaways
- One NETINT VPU encodes up to 32 1080p30 streams and decodes 48 streams while drawing under 20 watts
- A 1RU server with 10 VPUs delivers 320 encoded streams, 480 decoded streams, and 120+ encoding ladders across H.264, HEVC, and AV1
- Data-center GPU lead times run 36–52 weeks; DRAM lead times exceed 40 weeks, per the article
- The strategy frees new server purchases for AI and compute-intensive workloads rather than video transcoding
Why It Matters
The immediate implication is that organizations with retired servers can add AV1, HEVC, and H.264 encoding capacity without new capital expenditure or waiting on 36–52 week GPU lead times. For the broader streaming ecosystem, this reframes the infrastructure conversation: video encoding does not need high-performance CPUs when purpose-built silicon handles the workload, freeing new server purchases for AI and compute-intensive tasks. Watch whether infrastructure teams begin explicitly separating their encoding hardware budgets from their AI compute budgets — and whether VPU deployment density on recommissioned servers becomes a tracked capacity metric.
Additional Context
NETINT's recommissioning pitch builds on momentum from its September 2024 partnership with Akamai, which made VPUs available in the cloud for the first time. Akamai launched Accelerated Compute Instances powered by NETINT Quadra T1U VPUs in March 2025, starting at $0.42 per hour, and expanded to an 8-card VPU plan in April 2026 capable of 320+ parallel streams, per Akamai's product blog. NETINT reports more than 200,000 VPUs in production, with published benchmarks from independent test lab Cires21 showing 4–6x energy efficiency improvements over GPU alternatives at equivalent VMAF quality. The supply chain pressure NETINT cites is well-documented. Per GPUaaS.com's May 2026 availability report, NVIDIA B200 SXM lead times for enterprise buyers with OEM relationships run 8–16 weeks, while non-priority procurement faces 30+ week waits. Silicon Analysts reported in June 2026 that TSMC's CoWoS advanced packaging capacity remains fully booked at 52–78 week lead times, with an estimated 85%+ of 2026–2027 capacity already pre-locked — primarily by NVIDIA. DRAM lead times have also extended significantly, consistent with the 40+ week figure NETINT references. The timing aligns with a decisive shift in codec and hardware strategies. NETINT's 2026 State of Video Encoding Report, based on 286 industry professionals and covered by StreamingMediaBlog in March 2026, found AV1 at 17% production deployment with 40% of respondents planning deployment in 2026 — a 231% planned growth rate. VPU evaluation intent reached 49%, nearly matching GPU evaluation intent at 51%, while GPU-specific pain points around power consumption (39%), codec gaps (37%), and insufficient stream density (35%) are driving hardware diversification. Separately, StreamingMedia Magazine's 2026 codec analysis noted that Netflix now serves roughly 30% of streams via AV1, and 88% of large-screen devices certified between 2021 and 2025 include AV1 hardware decode.
Read full article at netint.com
Get this in your inbox → Subscribe
Enjoy our coverage?
Add StreamingMeme as a preferred source on Google to see more of our streaming news at the top of your Search results.
Add as preferred source