This Could Kill NVIDIA

This Could Kill NVIDIA

by Anastasi In Tech · View the channel on PSI

Applied Sciences & Engineering Economics & Finance KNTMedia, entertainment, information and communication industriesKNTXInformation technology industries
🎙 Anastasi In Tech 👥 505K 📅 September 30, 2026 ⏱ 14 min 👁 211K 📄 expert opinion 🧭 2026-10-06
Available in: English (current) Français

Keywords

interconnectphotonicsplasmonicsterahertzdata center

Summary

The video opens with the striking image of a Nevada data center filled with powerful GPUs that spend most of their time waiting for data. Anastasi In Tech argues that as AI models grow, the bottleneck in computing has shifted from raw chip speed to the speed and efficiency of interconnects between chips. She explains how transistors defined progress for sixty years, but now modulators and optical links are becoming the critical components. She introduces photonics, highlighting its enormous bandwidth advantage over electrical links, and then explains the trade-offs of Mach-Zehnder and micro-ring modulators. She presents plasmonics as a way to shrink modulators to nanoscale by coupling light with electrons, and introduces terahertz technology as a third option that avoids optical components altogether. Anastasi reveals that NVIDIA is investing billions in seven companies working on these technologies, and discusses NVLink Fusion, NVIDIA’s strategy to make its interconnect architecture essential even if customers use custom accelerators. She concludes by tying these advances to real-world costs, such as electricity consumption per query, and positions the network as the new determinant of AI performance and economics.

183 words

Critical Evaluation

Value of the Information & Strength of the Argument

The video offers substantial value by clearly demystifying a complex topic: why interconnects are now the primary performance limiter in AI systems. Anastasi effectively explains the transition from electrical to optical communication, and introduces newer concepts like plasmonics and sub-terahertz links in an accessible manner. Her argumentation is logically coherent: she establishes the problem (GPUs idle waiting), traces the historical role of transistors, and then builds a case for why modulators and optical interconnects are the new choke point. She supports her thesis by citing NVIDIA’s investment pattern and her own startup’s work on terahertz technology, which adds a practical dimension. However, the argumentation is partly self-promotional, and the claim that these technologies could ‘kill’ NVIDIA is exaggerated—she actually shows NVIDIA adapting. Nevertheless, the core reasoning about the shifting value from compute to connectivity is persuasive and well-illustrated.

Scientific Rigor, Source Quality, Title Accuracy

The scientific rigor is limited by the lack of specific citations or direct references to research papers, technical reports, or official announcements. Anastasi mentions NVIDIA investments in ‘seven companies’ without naming them, and references Polariton Technologies but provides no links. She relies heavily on her own expertise and the general state of the art, which is credible but not verifiable from the video alone. The title ‘This Could Kill NVIDIA’ is sensationalist and not fully supported by the content, which more accurately describes how NVIDIA is preemptively investing in interconnect technology to maintain dominance. The promotional segment for a GPT-6 Astra crash course, lasting roughly one minute, is clearly an advertisement and does not align with the scientific content. Overall, the video is informative and technically accurate, but its sourcing is weak and the title overstates the threat to NVIDIA.

293 words

Title / Content Match

The title suggests an existential threat to NVIDIA, but the video actually shows NVIDIA's proactive investments in interconnect technologies, presenting them as opportunities rather than threats. The mismatch is moderate, as the video does discuss potential shifts in value creation.

Quality & Reliability

7/10

The video provides a clear technical explanation of interconnect bottlenecks in AI data centers, but relies on the author's expertise and mentions investments without concrete references. The promotional segment for an AI crash course slightly detracts from the overall reliability.

Key Moments

Markers derived by PSI from the transcript: the creator did not define chapters.

Cited Sources

Contribution & Novelties

The video contributes by framing interconnects as the new ’transistor’ of the AGI era, a perspective that shifts attention from raw compute to networking. It introduces plasmonics and sub-terahertz as emerging solutions, and explains NVIDIA’s strategic pivot to interconnect architecture. The author’s background in chip design lends credibility to the technical explanations, and her own startup offers a concrete example of the terahertz approach. The video synthesizes these ideas in an accessible way, making cutting-edge technology understandable to a broader audience.

Pour aller plus loin :

  • Photonic integrated circuit — Relevant to understanding how optical components are integrated on chips.
  • Plasmonics — Directly tied to the plasmonic modulators discussed in the video.
  • Terahertz communication — Explores the sub-terahertz and terahertz band for data transmission.
  • NVLink — NVIDIA’s interconnect technology, central to the discussion of network architecture.

136 words

Radar Profile

The radar profile shows a balanced video with strong information quantity and adequate technical depth, while reliability is slightly lower due to the lack of external citations. The overall shape indicates a content that is informative and expert-driven but could benefit from more verifiable references.

Reliability 6/10

💬 très positif. Sur les 30 commentaires analysés, la majorité exprime une grande admiration pour la passion et la clarté de la présentatrice, avec quelques remarques techniques et humoristiques, sans critique négative significative.