Skip to content
Catalyst Note CN markCNCatalystNote

Noting the changes that set others in motion.

AI hardware

DeepX claims AI chip runs 40C cooler than rivals in butter demo

South Korean startup DeepX demonstrates thermal efficiency of its DX-M1 processor using a dairy-based benchmark against unnamed competitors.

DeepX claims AI chip runs 40C cooler than rivals in butter demo
Depth
1 outlet · newest source 164h ago

Key points

  • DeepX CEO cites a 25 to 40 degree Celsius temperature gap versus rival chips.
  • The DX-M1 processor is manufactured on Samsung 5nm process technology.
  • Mass production for the next-generation DX-M2 chip targets the second half of 2027.

South Korean startup DeepX claims its DX-M1 AI processor runs significantly cooler than competing chips, a capability it demonstrated using a block of butter as a thermal indicator. According to Tom's Hardware, the company presented this dairy-based benchmark to highlight efficiency differences in data center hardware.DeepX uses a butter melting test to show its AI chip runs much cooler than others.DeepX asserts its DX-M1 NPU achieves a 25-40C thermal delta over competitors under identical AI workloads.

Evidence: reportedBizarre AI chip demo uses butter to highlight how new chip runs 40 degrees cooler than the competition

aifleet-tomshardware

Evidence: reportedBizarre AI chip demo uses butter to highlight how new chip runs 40 degrees cooler than the competition

aifleet-tomshardware

The claim

DeepX CEO Kim Nok-won stated that the DX-M1 NPU maintains lower temperatures than unnamed rival processors while handling identical AI workloads. In a side-by-side demonstration reported by Tom's Hardware, competitor chips caused the butter to melt quickly, whereas the butter atop the DeepX solution remained solid. Kim asserted a temperature difference of 25 to more than 40 degrees Celsius compared with competitors.

Evidence: reportedBizarre AI chip demo uses butter to highlight how new chip runs 40 degrees cooler than the competition

aifleet-tomshardware

Evidence: reportedBizarre AI chip demo uses butter to highlight how new chip runs 40 degrees cooler than the competition

aifleet-tomshardware

Evidence: reportedBizarre AI chip demo uses butter to highlight how new chip runs 40 degrees cooler than the competition

aifleet-tomshardware

The response

Tom's Hardware noted significant gaps in the comparison data, specifically questioning the processing speed of each chip during the test. The outlet observed that any processor can run cooler if power and clock speeds are reduced, a practice common among PC enthusiasts who undervolt hardware for better thermals and reduced noise. This raises uncertainty about whether the cooling advantage comes from efficiency or reduced performance.

Why this matters now

The demonstration addresses growing global concerns about AI data center proliferation and their environmental impact. CNBC reported that hot-running processors drive extensive power and cooling apparatus needs, shaping business, politics and environmental outcomes.reported DeepX positions its design as a direct answer to these thermal challenges by making the chips themselves run cooler.

What buyers should weigh

Enterprise IT buyers should evaluate the trade-off between thermal efficiency and computational throughput. While lower operating temperatures can reduce cooling infrastructure costs, the lack of performance metrics in the demo leaves the actual AI workload capacity unclear. Buyers must verify if the cooling advantage stems from architectural efficiency or throttled performance. Additionally, the DX-M1 is produced on Samsung 5nm, with orders reportedly reaching $13.3 million, indicating early market validation despite the unconventional testing method.

Evidence: reportedBizarre AI chip demo uses butter to highlight how new chip runs 40 degrees cooler than the competition

aifleet-tomshardware

Why it matters

  • Lower chip temperatures could reduce data center cooling infrastructure costs and energy consumption for enterprise deployments.
  • Unverified performance metrics in thermal demos create risk for buyers prioritizing raw AI compute speed over efficiency.
  • Early orders of $13.3 million for the DX-M1 suggest initial vendor interest in alternative thermal solutions.
  • The planned H2 2027 mass production of the DX-M2 on Samsung 2nm indicates a roadmap for future efficiency gains.

Topics: AI chips, data centers, thermal management, DeepX, Samsung

Updated

Then / now / next

Then

Data center designs relied on extensive cooling apparatus for hot-running AI processors.

Now

DeepX demonstrates a processor that keeps butter solid during identical AI workloads.

Next

Buyers will watch for mass production of the DX-M2 chip in the second half of 2027.

Story timeline

  1. Then Data center designs relied on extensive cooling apparatus for hot-running AI processors.
  2. 09 Nov 2026 Bizarre AI chip demo uses butter to highlight how new chip runs 40 degrees cooler than the competition
  3. Now DeepX demonstrates a processor that keeps butter solid during identical AI workloads.
  4. Next Buyers will watch for mass production of the DX-M2 chip in the second half of 2027.

The strongest doubt

Counterpoint from the sources

Tom's Hardware questioned the validity of the thermal comparison, noting that reduced clock speeds can artificially lower temperatures without improving true efficiency.

For readers

Business

  • Cooler chips may reduce data center cooling infrastructure costs
  • Early orders of $13.3 million indicate market interest in the DX-M1

Investors

  • DeepX plans next-gen DX-M2 chip mass production for H2 2027
  • Company utilizes Samsung 5nm and future 2nm process nodes

Researchers

  • Dairy-based thermal benchmarks highlight efficiency differences
  • Undervolting practices complicate direct thermal performance comparisons

Ask the article

Local keyword search of this article; no generated answers.

Did this article give you the evidence you needed?

No vote saved on this device.
Your choice stays on this device. No public tally.

Terms in this story

NPU

Neural Processing Unit, a specialized processor for AI workloads

Undervolt

Reducing voltage supplied to a chip to improve efficiency and thermals

Samsung 5nm

A semiconductor manufacturing process node used for chip production

Questions readers ask

How much cooler does DeepX claim its chip runs compared to competitors?

DeepX CEO Kim Nok-won claims a temperature difference of 25 to more than 40 degrees Celsius.

What process node is used for the DX-M1 processor?

The DX-M1 processor is produced on Samsung 5nm technology, according to The Elec reports.

When is the next-generation DX-M2 chip expected to enter mass production?

DeepX aims to begin mass production of the DX-M2 chip in the second half of 2027.

Connected stories

This storyNvidia Board Meets to Discuss CEO Jensen Huang’s Future Amid Spy ScandalNvidia Board Meets to Discuss CETreasury chief blames Commerce for Nvidia chip export delays amid China scandalTreasury chief blames Commerce fNvidia Market Value Drops Over $1 Trillion Amid CEO Scandal and ResignationNvidia Market Value Drops Over $US Election Results Signal Stricter Data Center Rules for AI InfrastructureUS Election Results Signal Stric

Reading receipt

  • DeepX CEO cites a 25 to 40 degree Celsius temperature gap versus rival chips.
  • The DX-M1 processor is manufactured on Samsung 5nm process technology.
  • Mass production for the next-generation DX-M2 chip targets the second half of 2027.

Still open: What are the specific clock speeds and power limits used during the butter demonstration?

Still open: How does the DX-M1's AI inference latency compare to the rival chips that melted the butter?

Time spent: 0:00 · Depth: Full

Sources

  1. Tom's Hardware

Related stories

Background reading