Why We’re Heading back to Supercomputing 2026
AI and high-performance computing (HPC) are still pushing hardware refresh cycles faster. This year, something else has changed: the hardware coming out of the rack may be worth more than the day it was scheduled for replacement.
That’s why we, at Inteleca, are heading to Supercomputing 2026 (SC26) in Chicago this November, to talk about both ends of the HPC lifecycle: recovering value from retiring hardware and supplying proven infrastructure to teams that need capacity now.
Key Takeaways
- SC26 runs November 15 to 20, 2026 at McCormick Place in Chicago, with exhibits open November 17 to 19.
- Memory, storage, CPU and GPU prices are well above 2025 levels, driven by AI demand and tight supply.
- Retiring HPC hardware may be worth more today than when its replacement was first planned.
- Much of that value now sits in components like DIMMs and drives, so a part-level inventory matters.
- Flash and NVMe storage need purge-level erasure aligned with NIST SP 800-88 Rev. 2 and IEEE 2883, not a basic overwrite.
- Inteleca is attending SC26 to help organizations access GPU infrastructure without long lead times while maximizing the value of their existing hardware during infrastructure refreshes.
What is Supercomputing 2026
SC26, the International Conference for High Performance Computing, Networking, Storage, and Analysis, runs November 15 to 20, 2026 at McCormick Place in Chicago. The exhibit floor is open November 17 to 19.
This year’s theme is “HPC Unites,” bringing together national labs, universities, hyperscalers and hardware suppliers. The conference is expected to draw 14,000 to 18,000 attendees.
Read why we attended last year in our SC25 recap.
The HPC Hardware Market Right Now
Last year’s story was faster refresh cycles. In 2026, the components themselves have become scarce and expensive.
Record Memory and Storage Prices
AI build-outs are absorbing most of the world’s memory supply. TrendForce expected server DRAM prices to jump around 90% in a single quarter in early 2026, a record increase, and NAND flash contract prices up 70 to 75% in the second quarter.
Still Climbing, With No Quick Fix
TrendForce expects DRAM up another 10 to 15% and NAND up 15 to 20% in the fourth quarter, with meaningful new capacity not until late 2027 or 2028. Lead times on new systems can stretch to 40 weeks.
The result is an unusual window: retiring hardware is in high demand, and teams that need capacity can’t always wait for new systems.
Why SC26 Matters to Inteleca This Year
“SC26 is the one week a year when the people who build, run, and eventually retire the world’s biggest compute clusters are all in the same building. We help organizations recover value from the gear they’re retiring, and we supply proven infrastructure to teams that need capacity now and can’t wait out a 40-week lead time. The refresh cycle has never moved this fast, so we want to be in those conversations early, not after the hardware is already sitting in a cage.”
Paul Hogg, CEO, Inteleca
Retiring HPC Hardware in a Seller’s Market
“The market is upside down right now. Memory, storage, CPUs and GPUs are all selling for more than they did six months ago, and well above 2025 prices. So the hardware you’re retiring may be worth more today than the day you planned to replace it. Markets like this don’t last forever. The organizations that come out ahead are the ones with a decommissioning plan ready when the window is open.”
Paul Hogg, CEO, Inteleca
Here’s what HPC and data center teams should plan for:
- Know what’s in your racks. Track DIMMs, drives, CPUs and GPUs individually, since much of the value now sits in the parts.
- Use the right erasure for flash. A basic overwrite isn’t enough for NVMe and flash. Use purge-level erasure aligned with NIST SP 800-88 Rev. 2 and IEEE 2883, with a certificate for every drive.
- Plan ahead. A ready decommissioning plan lets you act while prices are high.
- Choose a certified partner. Our R2v3 certification covers data security, worker safety, environmental responsibility and downstream recycling.
What We’re Hearing From Customers
Across HPC and data center teams, a few themes keep coming up this year:
- Shorter refresh cycles. AI hardware now turns over in as little as 12 to 18 months, so ITAD is becoming part of infrastructure planning instead of an afterthought.
- More secondary-market sourcing. With long lead times on new systems, more buyers are sourcing tested hardware from the secondary market to get capacity online sooner.
- Earlier data security questions. Flash-heavy systems mean teams want an erasure plan in place before decommissioning starts, not after.
That’s the conversation we want to have at SC26: before the next refresh, while there’s still time to plan for value recovery, redeployment and secure erasure.
Our Focus at SC26: Both Ends of the HPC Lifecycle
Procurement → Deployment → Utilization → Retirement → IT Asset Disposition (ITAD)
At SC26, our work centers on four things:
- Value recovery: Getting the most from retiring servers, GPUs, memory and storage through our IT asset disposition (ITAD) services.
- Proven infrastructure: Tested HPC and AI hardware for teams that can’t wait out long lead times. B200, H200, H100, and A100 are a few examples of the GPUs our clients typically need, but we help procure the infrastructure your organization needs with custom configurations available.
- Data security: Certified, purge-level erasure for NVMe and flash.
- Sustainability: Reuse, resale and responsible recycling to cut e-waste.
Heading to SC26 and planning a refresh or looking for GPU infrastructure?
Contact our team to set up a time to connect in Chicago.
FAQs
Why are memory and storage prices so high in 2026?
AI data center build-outs are absorbing most of the world’s DRAM and NAND supply. New capacity isn’t expected until late 2027 or 2028, so prices remain well above 2025 levels.
Is retired HPC hardware worth more right now?
Often, yes. Higher component prices raise the resale value of retired servers, GPUs, memory and drives. A planned ITAD program helps you capture that value while it lasts.
How should data be erased from NVMe and flash drives?
Use purge-level erasure aligned with NIST SP 800-88 Rev. 2 and IEEE 2883, with a certificate for every drive. A basic overwrite isn’t enough for flash. Learn more about reducing data security risks in ITAD.
How often are HPC and AI servers refreshed?
Many are now replaced every 3 to 5 years, and sooner in AI-heavy environments. That makes end-of-life planning part of HPC hardware procurement.
What should I look for in an HPC ITAD partner?
Look for certified data erasure, transparent value recovery and responsible recycling. R2v3 certification is a strong sign a provider meets these standards. For HPC hardware, finding a partner that knows the market value is critical to get the most return on your hardware.

