USA · operational · ai training

AICR

Also known as Artificial Intelligence Compute Resources (MGHPCC)

Phase 1 seed dataset, compiled by hand. These rows were built from public operator, laboratory and vendor sources. A mechanical second-reader pass has since fetched every cited source: 1092 of 1153 systems have a readable citation that names them, and 40 are genuinely weakly sourced. Every claim carries its source and a confidence tier. Treat anything below verified as a lead, not a citation.

Measured performance

Rmax: -
Rpeak: -
Rmax ÷ Rpeak: -
Cores: -
Power: -
Per watt: - GF/W

Figures are as last publicly reported for the configuration described below, not a live measurement. Where a system was upgraded in place, the post-upgrade configuration is shown and the earlier one appears in the timeline.

What this machine is made of

One row per supplier relationship. “Supplier at build” is the company that shipped the part at the time; where that company has since been acquired, the parent it rolls up to today is shown beside it. That distinction is what makes ticker-level aggregation possible across a thirty-year dataset.

Role	Part	Supplier at build	Quantity	Confidence	Source
Integrator	- Cambridge Computer in partnership with Dell Technologies and VAST Data was selected as hardware provider	Dell Technologies DELLNYSE	-	Reported	mghpcc.org
Accelerator	NVIDIA B200 SXM Accelerators CUDA (Blackwell) · TSMC 4NP · 1,000 W 248 NVIDIA B200 GPUs, 192 GiB HBM3e each	NVIDIA NVDANASDAQ	248 verified	Verified	docs.ycrc.yale.edu orcd.mit.edu
Accelerator	- 152 NVIDIA RTX Pro 6000 Blackwell GPUs, 96 GiB GDDR7 each	NVIDIA NVDANASDAQ	152 verified	Verified	docs.ycrc.yale.edu orcd.mit.edu
CPU	AMD EPYC 9575F CPUs 64C · 3.3 GHz AMD EPYC 9575F CPUs, 128 cores per node across two sockets	AMD AMDNASDAQ	-	Reported	docs.ycrc.yale.edu
Storage	VAST Data Platform Storage Disaggregated shared-everything VAST Data storage platform, 15 PiB of AI workload working storage	VAST Data	-	Reported	mghpcc.org orcd.mit.edu

AI datacenter metrics

Figures beyond an accelerator count, each with who stated it. What can and cannot be compared across AI datacenters.

Metric	Value	As of	Basis	Note	Source
Accelerators	400 accelerators	2026-10	Operator stated	248 B200 plus 152 RTX Pro 6000 Blackwell GPUs at initial build, per Yale and MIT; Early Access period.	docs.ycrc.yale.edu orcd.mit.edu

The read

The Massachusetts AI Hub's Artificial Intelligence Compute Resources environment at the MGHPCC in Holyoke, funded by a $31 million state grant announced in May 2025 and matched by the six MGHPCC member universities (Boston University, Harvard, MIT, Northeastern, UMass and Yale). The MGHPCC release of 17 December 2025 says Cambridge Computer, Dell Technologies and VAST Data were selected as hardware partners and that the site will hold hundreds of NVIDIA B200 and RTX GPUs. Yale's research computing documentation states AICR initially contains 248 B200 and 152 RTX Pro 6000 Blackwell GPUs on AMD EPYC 9575F nodes (GPU nodes with 2.2 TB of RAM) and is in an Early Access period; MIT's ORCD newsletter of December 2025 independently gives the same 248 and 152 GPU counts plus 15 PiB of AI working storage. No FLOPS or power figure is published in the sources fetched.

Timeline

2025-05 Announced source

Site & facility

MGHPCC Holyoke

Location: Holyoke, MA, USA

kempnerinstitute.harvard.edu news.harvard.edu

Change history

2026-10-07 Added AICR (MGHPCC)

Source check

We fetched this system's own citations and recorded whether each page actually mentions it. This is a corroboration signal, not a fact check, and it is published so you can see how well the sourcing holds up rather than take it on trust.

Cited source	Result	Found on the page
docs.ycrc.yale.edu	names system, part and count	NVIDIA B200 SXM · counts: 248
orcd.mit.edu	names system, part and count	NVIDIA B200 SXM · counts: 248

Checked 2026-10-09 by pnpm verify. Re-run it and the table changes with the web.