Japan's Insane Superchip Made Nvidia Panic
Published Nov 17, 2026 · 19:34
Likes: 3Avg views/h: 155
Its recent uploads
#5 of its last 5 · its usual 7K
Its uploads we’ve seen, oldest left, by views (log scale). Dashed: its usual.
Performance in numbers
Figures as of Nov 18, 2026, 12:25 AM UTC.
Japan's Insane Superchip Made Nvidia Panic by Engineered in Asia has 348 views, 3 hours after it was published on November 17, 2026.
It is getting 155 views an hour now, 1.9× what Engineered in Asia’s uploads usually get (7K).
| Views | 348 | |
|---|---|---|
| Views per hour now | 155 | |
| Times its channel’s usual | 1.9× | |
| Its channel’s usual per upload | 7K | 7,198 |
| Best day | Nov 18 · 348 | 348 |
| Likes | 3 | |
| Published | November 17, 2026 | |
| Length | 19:34 | |
| Category | Science & Technology | |
| Channel | Engineered in Asia | |
| Channel subscribers | 13K |
Views
348
Not enough reads yet.
Not enough reads yet.
Views/h now
155
Not enough reads yet.
× usual
1.9×
its usual 7K
Not enough reads yet.
Views per day
348
best day, Nov 18
Where it charts
Not on any chart we track right now.
From the YouTube record
Fields of the video as the YouTube API returns them.
- Published
- Nov 17, 2026, 10:10 PM
- Duration
- 19:34
- Category
- Science & Technology
- Audio language
- English (en)
- Title language
- English (en)
- Captions
- No
- Definition
- HD
- Format
- 2D
- License
- Standard YouTube
- Licensed content
- Yes
- Embeddable
- Yes
- Made for kids
- No
- Age-restricted
- No
- Paid promotion
- No
- Public stats
- Yes
Topics
- Technology
Title and description in 14 languages
- Arabic
- Bangla
- English
- Hebrew
- Japanese
- Korean
- Malayalam
- Punjabi
- Polish
- Russian
- Tamil
- Telugu
- Ukrainian
- Dutch
Description
Can Tokyo startup Preferred Networks crush Nvidia's AI chip monopoly with a superchip running inference 10x faster? While Nvidia commands a $5 trillion market cap, PFN's radical design bypasses sold-out HBM memory. - Preferred Networks reveals the L1000 and L2000 architectures, stacking standard memory directly on math silicon to destroy data transfer latency. - Nvidia brings in $96 billion in quarterly revenue, yet remains trapped waiting on scarce HBM supplies from Samsung and SK Hynix. - Japan's defense agency ATLA backs PFN through 2028 to secure domestic artificial intelligence against skyrocketing grid power shortages. - Nvidia drops an estimated $20 billion on Grok's designs and talent, proving the industry's biggest player fears the exact same memory bottleneck. - Graphcore collapses into a $600 million SoftBank buyout, warning that hardware breakthroughs fail without massive software ecosystems. AI inference is no longer limited by raw compute power; it is choked by the hallway between memory and processing cores. If PFN and Samsung can commercialize 3D stacked silicon by 2027, data center power demands will plummet and the economics of large language models will shift permanently. Can Preferred Networks break Nvidia's CUDA software moat, or will PFN face the same fate as Graphcore? Share your prediction in the comments below, like the video, and subscribe for inside breakdowns of the global semiconductor wars. 00:00 The Tokyo Superchip 01:34 Nvidia AI Dominance 03:11 The Memory Bottleneck 04:29 From TSMC to MN3 07:35 Eliminating The Hallway 09:15 The L1000 Architecture 11:05 Defense Backing and Samsung 13:07 The Fall of Graphcore 14:39 The Earth Simulator Precedent 15:30 Nvidias Grok Deal 17:58 The Inference Race #Nvidia #AIHardware #Semiconductors #PreferredNetworks #AIInference #TechNews #SamsungShow all
Can Tokyo startup Preferred Networks crush Nvidia's AI chip monopoly with a superchip running inference 10x faster? While Nvidia commands a $5 trillion market cap, PFN's radical design bypasses sold-out HBM memory. - Preferred Networks reveals the L1000 and L2000 architectures, stacking standard memory directly on math silicon to destroy data transfer latency. - Nvidia brings in $96 billion in quarterly revenue, yet remains trapped waiting on scarce HBM supplies from Samsung and SK Hynix. - Japan's defense agency ATLA backs PFN through 2028 to secure domestic artificial intelligence against skyrocketing grid power shortages. - Nvidia drops an estimated $20 billion on Grok's designs and talent, proving the industry's biggest player fears the exact same memory bottleneck. - Graphcore collapses into a $600 million SoftBank buyout, warning that hardware breakthroughs fail without massive software ecosystems. AI inference is no longer limited by raw compute power; it is choked by the hallway between memory and processing cores. If PFN and Samsung can commercialize 3D stacked silicon by 2027, data center power demands will plummet and the economics of large language models will shift permanently. Can Preferred Networks break Nvidia's CUDA software moat, or will PFN face the same fate as Graphcore? Share your prediction in the comments below, like the video, and subscribe for inside breakdowns of the global semiconductor wars. 00:00 The Tokyo Superchip 01:34 Nvidia AI Dominance 03:11 The Memory Bottleneck 04:29 From TSMC to MN3 07:35 Eliminating The Hallway 09:15 The L1000 Architecture 11:05 Defense Backing and Samsung 13:07 The Fall of Graphcore 14:39 The Earth Simulator Precedent 15:30 Nvidias Grok Deal 17:58 The Inference Race #Nvidia #AIHardware #Semiconductors #PreferredNetworks #AIInference #TechNews #Samsung
◆ Calculated by viralshooter · all other figures from the YouTube API