Subscribe
Sign in
Home
About Austin
Paid Subscription
Interviews
TV
Disclosure
Optics & Networking
GPUs, XPUs, CPUs
Foundry
Memory
Autonomy / Physical AI
Chiplets & Packaging
Edge AI & Devices
Big Tech & Cloud
Fundamentals
Transcripts
Archive
Will ASML Finally Charge for Its Value?
Management hints ASML starts charging for value in 2027. Consensus says it won't.
Jul 31
•
Austin Lyons
17
2
TSMC could spend $10B more on N2 capacity. Why won’t they?
TSMC has the cash... Why not build more N2? Implications of not doing so? And more
Jul 24
•
Austin Lyons
21
3
🎙️ PicoJool’s Al Yuen: The Case for GaAs VCSELs in Scale-Up Interconnects
GaAs VCSELs at 200G/lane, unconstrained supply vs. indium phosphide, the 8×200 / 16×100 / 32×50 flavors, WIN Semiconductors, the roadmap to 12.8T, and…
Jul 16
•
Austin Lyons
14
1
Most Popular
View all
Advanced Packaging: Intel's EMIB vs TSMC's CoWoS
May 12
•
Austin Lyons
143
19
GPU Networking Basics, Part 1
Mar 19, 2025
•
Austin Lyons
219
29
GPU Networking Basics, Part 2
Apr 10, 2025
•
Austin Lyons
67
11
Credo AECs
Sep 22, 2025
•
Austin Lyons
64
12
GPU Networking Basics Part 3: Scale-Out
Sep 9, 2025
•
Austin Lyons
53
15
SerDes Matters!
Oct 22, 2025
•
Austin Lyons
49
5
Recent posts
View all
The Next Trillion-Dollar Chip Company
It might be simpler than you think. Etched, Fractile, MatX, Positron, Rebellions, SambaNova, Tensordyne, Tenstorrent, and more.
Jul 15
•
Austin Lyons
40
3
High Bandwidth Flash: The Full Report
Why HBF for inference decode fits, what it does to the DRAM and NAND markets, a worked cost model, who moves first, and risks
Jul 7
•
Austin Lyons
25
6
Micron's Blowout and the Case for a Memory Supercycle
Micron's GM hit 84.9%, it booked $10B of take-or-pay through 2030, and supply shortage stretched further out. Peak? When does supply increase? SK Hynix…
Jun 26
•
Austin Lyons
19
1
The ISA Doesn't Matter Where It Counts
x86's lock-in is real. It's just furthest from the GPU.
Jun 18
•
Austin Lyons
20
1
Are Agentic CPUs a Commodity? It’s Complicated.
Five sockets, different economics. A model for telling them apart, and where each competitor fits. AMD, Intel, Nvidia, Arm, Qualcomm.
Jun 10
•
Austin Lyons
34
4
Interviews
View all
🎙️ PicoJool’s Al Yuen: The Case for GaAs VCSELs in Scale-Up Interconnects
GaAs VCSELs at 200G/lane, unconstrained supply vs. indium phosphide, the 8×200 / 16×100 / 32×50 flavors, WIN Semiconductors, the roadmap to 12.8T, and…
Jul 16
•
Austin Lyons
14
1
44:47
An Interview with Nvidia's Deepu Talla About Physical AI and Robotics
Industrial businesses orchestrating across embodied & digital agents, world models, hybrid edge-cloud & "phone a friend," Nvidia Mega, Newton physics…
May 25
•
Austin Lyons
22
An Interview with the Gimlet Labs Team About Heterogeneous Inference for AI Agents
Why most neoclouds can't follow Gimlet's silicon-vendor-neutral model, d-Matrix Corsair + NVIDIA B200 delivering 4× Pareto frontier shifts on GPT-OSS…
May 12
•
Austin Lyons
21
2
An Interview with Meta VP Matt Steiner About Ads Infrastructure
MTIA, co-designed NVIDIA SKUs, LLM-written kernels, a 1T-parameter recommender at sub-second, and more
Apr 20
•
Austin Lyons
14
2
Optics & Networking
View all
🎙️ PicoJool’s Al Yuen: The Case for GaAs VCSELs in Scale-Up Interconnects
GaAs VCSELs at 200G/lane, unconstrained supply vs. indium phosphide, the 8×200 / 16×100 / 32×50 flavors, WIN Semiconductors, the roadmap to 12.8T, and…
Jul 16
•
Austin Lyons
14
1
Inside the 800G → 1.6T → 3.2T Race
Timing is everything. What the industry said about 800G, 1.6T, 3.2T in recent earnings calls
May 19
•
Austin Lyons
24
1
VCSELs + 200G Wall In AI Datacenters?
Decades of short-reach dominance, the supply chain holding it up today, and the problems at the 200G transition
May 6
•
Austin Lyons
23
1
Coherent's Vertical Integration Strategy
Coherent makes more of the optical stack in-house than any competitor. We walk through the business, the growth vectors, and whether breadth beats…
Apr 1
•
Austin Lyons
27
4
GPUs / XPUs / CPUs
View all
The Next Trillion-Dollar Chip Company
It might be simpler than you think. Etched, Fractile, MatX, Positron, Rebellions, SambaNova, Tensordyne, Tenstorrent, and more.
Jul 15
•
Austin Lyons
40
3
The ISA Doesn't Matter Where It Counts
x86's lock-in is real. It's just furthest from the GPU.
Jun 18
•
Austin Lyons
20
1
Are Agentic CPUs a Commodity? It’s Complicated.
Five sockets, different economics. A model for telling them apart, and where each competitor fits. AMD, Intel, Nvidia, Arm, Qualcomm.
Jun 10
•
Austin Lyons
34
4
An Interview with the Gimlet Labs Team About Heterogeneous Inference for AI Agents
Why most neoclouds can't follow Gimlet's silicon-vendor-neutral model, d-Matrix Corsair + NVIDIA B200 delivering 4× Pareto frontier shifts on GPT-OSS…
May 12
•
Austin Lyons
21
2
Foundry
View all
Will ASML Finally Charge for Its Value?
Management hints ASML starts charging for value in 2027. Consensus says it won't.
Jul 31
•
Austin Lyons
17
2
TSMC could spend $10B more on N2 capacity. Why won’t they?
TSMC has the cash... Why not build more N2? Implications of not doing so? And more
Jul 24
•
Austin Lyons
21
3
Advanced Packaging: Intel's EMIB vs TSMC's CoWoS
Is Intel's EMIB better than TSMC's CoWoS for AI accelerators? A primer on both, an honest look at the trade-offs, and where it goes from here.
May 12
•
Austin Lyons
143
19
TSMC's Margins in Uncharted Territory
66.2% in the weakest quarter of the year. The biggest sequential jump on record. And 10 points above the long-term target they just raised.
Apr 23
•
Austin Lyons
5
1
Autonomy / Physical AI
View all
44:47
An Interview with Nvidia's Deepu Talla About Physical AI and Robotics
Industrial businesses orchestrating across embodied & digital agents, world models, hybrid edge-cloud & "phone a friend," Nvidia Mega, Newton physics…
May 25
•
Austin Lyons
22
An Interview with Rivian’s Mukund Chavan About RAP1
Rivian designed its own autonomy SoC. Why?
Mar 5
•
Austin Lyons
8
The Nvidia Business Hiding in Plain Sight
And it's bigger than Wall Street thinks
Feb 17
•
Austin Lyons
20
2
Why LIDAR Is Consolidating Now
Innoviz CEO Omer Keilaf on the L4 Gold Rush, the L3 Unlock, and Why the Music Has Stopped
Jan 22
•
Austin Lyons
11
3
Chiplets & Packaging
View all
Power Moves Into the Package. Empower, PowerLattice, and the IVR Socket
Why did ADI agree to pay $1.5B for Empower Semi? Because XPUs are about to draw 3,000+ amps at 0.7V. Transients and I²R both blow up. Move the regulator…
May 27
•
Austin Lyons
22
1
Advanced Packaging: Intel's EMIB vs TSMC's CoWoS
Is Intel's EMIB better than TSMC's CoWoS for AI accelerators? A primer on both, an honest look at the trade-offs, and where it goes from here.
May 12
•
Austin Lyons
143
19
An Interview With Mohamed Awad About Chiplets
Arm's Chiplet Ecosystem Announcement
Oct 15, 2025
•
Austin Lyons
8
1
Building the Chiplet Ecosystem
Arm’s CSA, Neoverse CSS, and Total Design as a framework for multi-vendor chiplet ecosystem alignment
Aug 20, 2025
•
Austin Lyons
21
6
Edge AI & Devices
View all
The Agentic Computer: New S-Curve or Another iPad?
The next device category is here. But will the economics work?
Apr 7
•
Austin Lyons
18
Intel's Product Marketing at CES: B-
Better grasp of reality, but yet again the AI pitch missed. PTL 18A looks great though!
Jan 13
•
Austin Lyons
7
1
When Frontier AI Goes From Cloud to Desk in Five Years
Implications of running yesterday's frontier models on today's local supercomputers. Rhymes with Minicomputers in Mainframe Era. Nvidia, AMD, Intel…
Jan 8
•
Austin Lyons
9
WWDC25: Apple’s Playing a Different Game
The strategy makes sense. The narrative is missing. Let’s examine Apple’s AI approach from first principles and assess whether WWDC delivered.
Jun 16, 2025
•
Austin Lyons
14
Big Tech & Cloud
View all
Rethinking Amazon + Not All SaaS Is Dead
AWS Shows a Survival Path for Enterprise Software
Apr 30
•
Austin Lyons
16
An Interview with Meta VP Matt Steiner About Ads Infrastructure
MTIA, co-designed NVIDIA SKUs, LLM-written kernels, a 1T-parameter recommender at sub-second, and more
Apr 20
•
Austin Lyons
14
2
Meta's MTIA Roadmap
Inference-centric design. ROIC is clear. Implications for the industry.
Mar 12
•
Austin Lyons
13
1
The $200 Billion Bet
Why Amazon has the hardest ROIC story in tech, and what to do about it
Feb 13
•
Austin Lyons
18
1
Fundamentals
View all
High Bandwidth Flash: The Full Report
Why HBF for inference decode fits, what it does to the DRAM and NAND markets, a worked cost model, who moves first, and risks
Jul 7
•
Austin Lyons
25
6
Optics Primer, Part 3: Co-Packaged Optics (CPO)
From EML lasers and DSPs to silicon photonics and external CW lasers. How CPO works and the impact on the optical supply chain.
Mar 9
•
Austin Lyons
30
1
Optics Primer, Part 2: LRO & LPO
LRO and LPO move signal processing out of the transceiver and into the switch, saving power at the cost of tighter system coupling. An incremental step…
Jan 26
•
Austin Lyons
33
5
LiDAR, Explained: How It Works and Why It Matters
Autonomy, 905nm vs 1550nm, ToF vs FMCW, market dynamics, and more.
Jan 19
•
Austin Lyons
14
4
This site requires JavaScript to run correctly. Please
turn on JavaScript
or unblock scripts
