Assertion Partly supported AI assessment confidence: 85% certainty 3/5 debate potential 2/5

Ross: Roughly 80% of Global HPC Servers Sit Powered and Idle

Mark Ross · EP 444: $1.5M in MRR, SaaS, Helping Studios Render Film Faster with GridMarkets CEO Mark Ross · Oct 11, 2016 · at 3:19

Mark Ross, CEO of GridMarkets, explains the market opportunity behind aggregating dark CPU compute capacity for visual effects and rendering.

0:00 / 0:12exact quote · 12.4s

720p mp4 · rendered on demand · StarZero watermark

“So today there are roughly 80% of the high performance computing servers that are in the world are not being used. They're idle. They're electrified. They're racked and stacked, but they're just sitting there waiting for something to come to them.”

quote is from the automated transcript, cleaned for reading: filler sounds and stutters are removed, nothing is rephrased. names can be misheard (the analysis reads context, assessments check outside sources). how →

The claim, plainly

Roughly 80% of high-performance computing servers globally sit powered, racked, and idle without being used.

a quote alone can be ambiguous, so this restates it with the references and context of the conversation around it resolved. AI-written; the tape below the player is the source.

Assessment

The claim is partially true. Industry benchmark studies published around 2014–2016 (including reports by the Natural Resources Defense Council, Lawrence Berkeley National Laboratory, and McKinsey & Company) documented that traditional enterprise data centers and on-premise burst computing clusters operated at average utilization rates of 10% to 18%, meaning roughly 82% to 90% of provisioned server capacity was idle at any given time. Private on-premise enterprise HPC clusters also averaged low utilization (20% to 35%) due to peak-load provisioning. However, dedicated high-performance computing (HPC) environments (such as national supercomputing labs and academic research clusters) generally operated under heavy batch queue scheduling with allocation rates exceeding 70% to 90%. The figure of ~80% idle accurately reflected the broader enterprise server baseline and burst enterprise HPC setups, but does not apply universally to dedicated HPC facilities.

2014 NRDC/Anthesis Group report found average enterprise data center server utilization was between 12% and 18%, leaving 82% to 88% of capacity idle.
2016 LBNL U.S. Data Center Energy Usage Report estimated operational server utilization in corporate data centers at 10% to 15% (85% to 90% unutilized).
2015 Stanford/Anthesis study by Koomey and Taylor identified that ~30% of enterprise servers were completely 'comatose' (zero compute output for 6+ months).
Dedicated academic and national lab HPC clusters historically maintain high scheduler allocation rates of 70% to 90%+, contrasting with enterprise burst clusters.

Checked against

[1] davidkopp.de [2] nrdc.org [3] devsustainability.com [4] storageconsortium.de [5] forbes.com [6] stanford.edu [7] anthesisgroup.com [8] optops.ai

"HPC" cluster utilization rate percentage "idle" average server utilization data center 10 15 20 percent idle NRDC server utilization rate Gartner IDC NRDC 2014 2015 2016 Anthesis comatose servers Jonathan Koomey 2015 2016

assessed as of 2026-09-02

This assessment is an AI reading of a public statement, made against the cited references with the confidence shown. It can be wrong. The quote, the tape and the references are all on this page; if something is off, use report an error below. full disclaimer →

Mentioned

GridMarkets high performance computing servers

⚑ Spot an error? Report it

More from Mark Ross

Assertion Not checkable as stated

Ross: GridMarkets cuts rendering turnaround from months to hours

“We serve studios who make movies, animations, and special effects, and they would use our service to submit what's called rendering and simulation, which is a very computationally intensive application, and through a button that they can push in their applicat…”

Mark Ross Oct 11, 2016 ▶ 2:35 EP 444: $1.5M in MRR, SaaS, Helping Studios Render Film Faster with GridMarkets CEO Mark Ross

Assertion Not checkable as stated

Ross: GridMarkets generates around $1.6 million in monthly revenue

“In that neighborhood. Yeah.”

Mark Ross Oct 11, 2016 ▶ 11:23 EP 444: $1.5M in MRR, SaaS, Helping Studios Render Film Faster with GridMarkets CEO Mark Ross

Disclosure

Ross: GridMarkets Requires No SLAs or Machine Commitments from Compute Suppliers

“It's an arrangement that really does not require a service level agreement with our suppliers. It doesn't require any commitment for them to provide us any, you know, specific amount of machines. The reason for that is we've built a grid, so if one supplier's …”

Mark Ross Oct 11, 2016 ▶ 4:09 EP 444: $1.5M in MRR, SaaS, Helping Studios Render Film Faster with GridMarkets CEO Mark Ross

Assertion Not checkable as stated

Ross: GridMarkets processes over 1M credit hours monthly

“Well, we've I would say we're in the probably just over the millions, million mark at the moment.”

Mark Ross Oct 11, 2016 ▶ 7:34 EP 444: $1.5M in MRR, SaaS, Helping Studios Render Film Faster with GridMarkets CEO Mark Ross

Assertion Not checkable as stated

Ross: GridMarkets currently serves over 700 customers

“We have over 700 at the moment.”

Mark Ross Oct 11, 2016 ▶ 10:41 EP 444: $1.5M in MRR, SaaS, Helping Studios Render Film Faster with GridMarkets CEO Mark Ross

Assertion Not checkable as stated

Ross: Rendering a standard 1-minute ad requires ~100 servers for 3–4 days

“A typical job will take probably around a hundred machines working potentially over a three to four day period. And that, that configuration would produce a typical advertisement. So, you know, maybe, maybe a minute long, or maybe a bit longer than that.”

Mark Ross Oct 11, 2016 ▶ 13:30 EP 444: $1.5M in MRR, SaaS, Helping Studios Render Film Faster with GridMarkets CEO Mark Ross

Ross: Roughly 80% of Global HPC Servers Sit Powered and Idle

More from Mark Ross

Turn any episode into a week of clips.