Test report DSG-1184 · Rev A · tested November 9, 2026
AI Datacenter InfrastructureDevice under test
Nvidia Invests in Point2 Technology at Seoul AI Forum
Nvidia invested an undisclosed sum in Point2 Technology, a Korean startup building RF interconnects for AI factories, at a Seoul forum on November 9, 2026.
- Read
- 4 min
- Words
- 735
- Node
- 7nm
- Operator
- Marcus Bennett
Spec summary
- Nvidia announced an investment in Point2 Technology on November 9, 2026 at COEX, Seoul.
- The investment amount, valuation and timeline were not disclosed.
- Nvidia named Motif Technologies, RLWRLD and Upstage as planned future investments.
- More than 20 Korean startups attended the forum hosted with the Ministry of Science and ICT.
- Point2 founder Bae Hyeon-min heads the startup institute at KAIST.
![Nvidia Backs Point2 at Seoul Forum: 20+ Startups [2026]](/media/3a/3a06c31f515b9394.1280.webp)
By Marcus Bennett
Nvidia invested in Point2 Technology, a South Korean startup building radio frequency interconnects for AI data centers, on November 9, 2026. The company did not disclose the amount. It announced the deal at the AI Startup Partnership Forum, hosted with Korea's Ministry of Science and ICT at COEX in Seoul.
More than 20 Korean startups attended the event. Nvidia also said further investments are planned in Motif Technologies, RLWRLD and Upstage. Those three names carry forward-looking language, not closed deals. No dollar figures, stake sizes or timelines appeared for any of the four companies.
What does Point2 Technology actually build?
Point2 develops scalable RF-based interconnects that move data between semiconductor chips and between racks in what Nvidia calls AI factories. Nvidia framed the approach as an alternative to conventional copper and optical interconnect solutions.
The mechanism is straightforward. An RF interconnect sends data on a radio frequency carrier wave over a guided channel, instead of pushing raw electrical pulses down copper traces or converting signals to light. The pitch: cover chip-to-chip and rack-level distances with less of the signal degradation that plagues fast copper.
Reporting identifies Bae Hyeon-min as a founder of Point2 and head of the startup institute at KAIST, one of Korea's leading engineering universities. That academic tie matters. Interconnect design draws on analog circuit and signal integrity research, fields where university labs often seed commercial teams.
What we do not have is just as important:
- No headcount, revenue, customer list or funding history
- No data rate, reach, energy-per-bit or latency figures
- No shipping or deployment timeline
Anyone quoting a valuation or ownership percentage is inventing it.
Why does the wiring matter?
Inside an AI factory, thousands of accelerators exchange data constantly. The links carrying that traffic decide how well a cluster scales, how much power it burns and how much it costs.
The incumbents both carry trade-offs. Copper is cheap, but electrical signals lose strength as speeds climb, which limits cable reach. Optical travels farther and carries more, but adds conversion hardware and more failure points. Engineers have searched for a middle path for years.
The pressure is real. Operators already struggle to find enough power and supply, and every watt spent moving bits is a watt unavailable for compute. Any interconnect that trims overhead gets a hearing.
How do RF, copper and optical compare?
On verifiable facts, the ledger is sparse. Copper and optical ship today, with public benchmark data widely available. Point2's RF approach has no disclosed benchmarks and no confirmed delivery date. Copper and optical are proven and bought at scale; RF is a promising entrant without public proof points.
Moving from slide to production rack takes qualification cycles that run long, and large operators adopt new links slowly. Nvidia's backing does not guarantee adoption, but it gives a small company credibility that money alone rarely buys.
What does this change for the market?
Immediate impact looks small. No amount was disclosed, Point2 is private, and public investors have nothing to model. The signal still reaches several groups.
Startups gain a template: a component maker can now point to Nvidia taking a stake at a government-hosted event. Data center operators gain a hint that the largest AI chip supplier is exploring options beyond copper and optical. Component makers across Asia gain a reason to sharpen their pitch.
The forum itself spanned four layers of the AI stack: foundation models, physical AI, AI infrastructure and AI semiconductors. Point2 straddles the last two. Its links serve infrastructure builders and touch the semiconductor layer because they connect chips.
What should teams watch next?
Infrastructure teams planning 2027 builds should keep current copper and optical plans intact. A startup with no public performance data is not yet a procurement option. Watch for:
- Published specifications from Point2 covering data rate, reach and power per bit
- Nvidia confirming deal terms for Motif Technologies, RLWRLD and Upstage
- Any cloud provider or system builder naming Point2 as a supplier
- Follow-on programs from the Ministry of Science and ICT for the 20-plus attending startups
The responsible reading is patience. Nvidia invested and described the technology. It did not say it will ship RF links inside a product. Until a filing, blog post or named executive supplies numbers, the amount stays undisclosed and the bet stays a vote of confidence.
via shattered.io (Source)
More from Marcus Bennett
Show full bio
News editor covering regulation, policy and their impact on hardware layer of artificial intelligence at Die Signal.
591 articles
Same lot · LOT-3610
- DSG-863110nmJensen Huang Exit Risk Splits Korean AI Startups on Stakes
- DSG-34013nmNVIDIA and Coherent Partner on Optics for Next-Gen Data Centers
- DSG-826414nmNvidia Invests in Data Center Developer Cloverleaf Infrastructure
- DSG-548565nmJensen Huang Signals NVIDIA's N1 Plans for AI PCs
- DSG-39483nmSK Group and NVIDIA Expand Partnership on AI Factories and Memory