TechSurge: Deep Tech Podcast

In this episode, host David Goldman speaks with legendary graphics chip architect Raja Koduri, who explains why every gigawatt of AI infrastructure now costs $50 to $60 billion, and why China's goal of doing it for under $10 billion is the real threat to Western AI. Raja twice led graphics at AMD, directed Apple's graphics architecture and was chief architect at Intel. He argues that the real AI race isn't Nvidia vs. Google vs. Broadcom but China vs. the rest of the world, and that the new bottleneck isn't compute. It's memory.

In this conversation, Raja joins TechSurge to talk about his new startup Oxmiq, which aims to turn "electrons to tokens super efficiently." He covers how 3D-stacked, hybrid-bonded memory could deliver 10x the bandwidth of today's HBM, why AI agents are changing how chips get designed, and why he thinks the next disruption to AI data centers "comes from the bottom."
The conversation covers:
✅ Why every gigawatt of AI infrastructure costs $50 to $60 billion, and where the $45 billion in hardware spend goes
✅ The $24 trillion capital question: 400+ gigawatts of new compute needed by 2030
✅ How China's under-$10 billion per gigawatt target creates a 5 to 6x cost gap
✅ Why memory hierarchy, not raw compute, is now the real bottleneck in AI
✅ How advanced packaging can unlock 10x bandwidth and 10x token generation, even on older 7nm nodes
✅ Why OpenAI's Jalapeño chip shows that AI can speed up silicon design
✅ Why the value of experienced engineers has gone up 100x in the age of AI coding agents
✅ Leadership lessons from Steve Jobs at Apple and Lisa Su at AMD
✅ Why Intel's decision to kill 3D XPoint memory came at "the wrong time"
✅ Boom or bust: the financial engineering risk behind the AI infrastructure buildout
✅ Token factories vs. token banks: why "the more boring you make it, the more it becomes fabulous"
Guest Links:
Raja Koduri: Founder of Oxmiq.
LinkedIn: https://www.linkedin.com/in/raja-koduri-3a51611
X: https://x.com/RajaXg
Oxmiq: https://oxmiq.ai 

Further Reading and Resources

OpenAI and Broadcom announcement: https://openai.com/index/openai-broadcom-jalapeno-inference-chip/

High Bandwidth Memory (HBM): The memory technology Raja's AMD team helped bring to market with HBM1 and HBM2, and the benchmark Oxmiq's 3D-stacked approach aims to beat by 10x. https://en.wikipedia.org/wiki/High_Bandwidth_Memory 
Intel 3D XPoint (Optane): The discontinued memory technology Raja says could have made Intel a major player in the inference era. https://en.wikipedia.org/wiki/3D_XPoint 
Chapters
00:00 - A Gigawatt of AI Now Costs $60 Billion
01:58 - Introducing Raja Koduri
02:02 - What Oxmiq Builds: Electrons In, Tokens Out
05:18 - The $24 Trillion AI Infrastructure Bill
06:05 - China vs. the Rest of the World
09:10 - Memory Is the New Bottleneck
22:03 - How AI Agents Are Changing Chip Design
36:30 - Lessons From Steve Jobs and Lisa Su
44:54 - Advanced Packaging, Memory Prices, and Intel's Mistake
55:38 - Boom or Bust: The Future of Token Factories

About TechSurge:

TechSurge Podcast shares the latest insights directly from legendary Silicon Valley leaders, daring new founders, and visionary technologists.
🔔 Subscribe for weekly conversations at the intersection of technology advancement, market dynamics, and founder journeys.

Almost 2% of U.S. GDP will be spent on AI infrastructure this year, nearly double 2025's figure. But beneath those headline numbers, the composition of that spending has quietly flipped: for the first time, dollars spent on running models in production now outweigh dollars spent training them.
 
In this episode of TechSurge, host David Goldman speaks with Austin Lyons, a semiconductor analyst at Creative Strategies, co-host of the Semi Doped podcast, and author of the Chipstrat newsletter. Lyons previously worked as a hardware engineer at Intel and as a product manager on John Deere's autonomous tractor and Blue River Technology teams before turning to full-time chip industry analysis.
 
The conversation opens with why AI buyers have moved from assembling commoditized parts to buying entire pre-integrated systems, tracing how Nvidia's rack-scale approach, exemplified by its 72-GPU Grace Blackwell racks, made turnkey deployment the default, and why that raises the bar for any chip startup trying to compete. Lyons and Goldman then unpack how inference workloads have split into two distinct problems, prefill and decode, and how that split created an opening for SRAM-based challengers to outperform general-purpose GPUs on decode speed.
 
From there, the discussion turns to the rise of neoclouds, the GPU-rental companies that grew into public businesses worth well over $100 billion combined, and why so many traditional investors missed them. Lyons and Goldman work through the circular financing debate head-on: the mechanics of Nvidia's equity stakes, GPU-backed debt, and hyperscaler off-take agreements that critics compare to dot-com-era vendor financing, and the counterargument that demand is simply outrunning fixed supply.
 
The episode closes on Lyons's own framework for identifying the next trillion-dollar chip company, built on four conditions including the ability to run trillion-parameter models at rack scale, beat an incumbent on a key performance metric, and land a frontier anchor customer, along with a look at how AI-assisted chip design is lowering the barrier for more companies, from OpenAI to electric vehicle makers, to design their own custom silicon.
 
Sign up for our newsletter at techsurgepodcast.com for updates on upcoming TechSurge Live Summits and future episodes.
 
Speaker Profiles and Links
 
David Goldman: Partner, Celesta Capital
 
Austin Lyons: Senior Analyst, Creative Strategies; Founder, Chipstrat; Co-host, Semi Doped
LinkedIn: https://www.linkedin.com/in/austinlyons/
Newsletter: https://www.chipstrat.com
 
Further Reading and Resources
 
Nvidia DGX GB Rack Scale Systems documentation: https://docs.nvidia.com/dgx/dgxgb200-user-guide/
OpenAI and Broadcom – "OpenAI and Broadcom Unveil LLM-Optimized Inference Chip": https://openai.com/index/openai-broadcom-jalapeno-inference-chip/
Chipstrat – Austin Lyons's newsletter: https://www.chipstrat.com
Semi-doped: https://semidoped.com/
 
Timestamps
 
00:00 — No One's Brought a Chip to Market Built for LLMs
01:21 — Introducing Austin Lyons
02:16 — Why AI Buyers Now Buy Whole Systems, Not Parts
08:18 — Nvidia's Margins and the Case for System Simplicity
10:10 — Can a Startup Compete When You Have to Sell Systems?
14:12 — Prefill vs. Decode: Splitting the Inference Workload
24:51 — Fragmentation vs. Consolidation in AI Silicon
28:22 — Why Investors Missed the First Wave of Neoclouds
38:18 — The Circular Financing Debate
48:29 — Lyons's Four Conditions for the Next Trillion-Dollar Chip Company
 
 About TechSurge:
TechSurge Podcast shares the latest insights directly from legendary Silicon Valley leaders,
daring new founders, and visionary technologists.
Subscribe for weekly conversations into the intersection of technology advancement, market dynamics, and founder journeys.
#AISilicon #LLMHardware #Nvidia #AIInference #TechPodcasts #AIInfrastructure

In this episode, Nobel Prize-winning physicist Dr. John Martinis reveals how his breakthrough in superconducting qubits made quantum physics real at macroscopic scale and what it means for the future of technology. The former lead of Google's Quantum AI lab explains why quantum computing is so fragile, why a lot of hype has a low chance to work, and why his fabless company Qolab could be the Nvidia of quantum computing
In this conversation, Dr. Martinis joins TechSurge to explain the science behind macroscopic quantum coherence, the engineering challenges of scaling quantum computers, and how hybrid quantum-classical computing will shape the future of technology.
The conversation covers:
✅ How the superconducting qubit breakthrough won the Nobel Prize in Physics
✅ Why Nature wants to destroy quantum coherence and why quantum is fragile
✅ From academic physics to building Google's quantum computer
✅ The engineering challenge of scaling quantum computing beyond the lab
✅ Why a lot of quantum computing hype has a low chance to work
✅ How Qolab's fabless model could scale quantum hardware
Guest Links:
John Martinis: 2025 Nobel Prize laureate in Physics, superconducting-qubit pioneer, former
Google quantum-hardware researcher, and founder and CTO of Qolab.
Nobel Prize profile: https://www.nobelprize.org/prizes/physics/2025/martinis/
Qolab: https://qolab.ai/
Further Reading and Resources
Google Sycamore Quantum Processor - Google’s 2019 experiment used a 53-qubit
superconducting processor to perform a specific random-circuit-sampling task substantially
faster than the then-known classical approach.
Nature research paper:
https://www.nature.com/articles/s41586-019-1666-5
Google Research explanation:
https://research.google/blog/quantum-supremacy-using-a-programmable-superconducting-processor/
Artificial Intelligence and Transformers – The Transformer architecture discussed in the
podcast was introduced in the paper “Attention Is All You Need.”
Original paper:
https://arxiv.org/abs/1706.03762
AlphaFold and Protein Structure Prediction – AlphaFold demonstrated how classical AI can
predict protein structures with high accuracy, illustrating the distinction between present-day AI
and potential future quantum applications.
Nature paper:
https://www.nature.com/articles/s41586-021-03819-2
Google DeepMind – AlphaFold:
https://deepmind.google/science/alphafold/
Quantum Computing Hardware Approaches – The podcast compares superconducting
qubits, semiconductor spin qubits, neutral atoms, trapped ions and photonic systems.
Google Quantum AI:
https://quantumai.google/
Intel Quantum Computing:
https://www.intel.com/content/www/us/en/research/quantum-computing.html
QuEra – Neutral-atom quantum computing:
https://www.quera.com/
Atom Computing:
https://atom-computing.com/
Quantum Manufacturing and Scaling – Qolab is focused on improving the fabrication, wiring and scalability of superconducting quantum processors through industrial partnerships.
Qolab:
https://qolab.ai/
Qolab and Applied Materials collaboration:
https://thequantuminsider.com/2025/03/18/qolab-secures-investment-from-applied-ventures-and-announces-collaboration-to-advance-quantum-computing-manufacturing/
Applied Materials:
https://www.appliedmaterials.com/
Quantum–Optical Networking – The podcast discusses the challenge of converting
microwave signals used by superconducting qubits into optical signals suitable for fiber-optic communication.
Microwave-to-optical conversion research:
https://www.nature.com/articles/s41567-019-0650-1
Chapters:
00:00 – The Quantum Computing Hype: Physics vs Engineering
04:06 – Introducing Nobel Prize Winner John Martinis
12:09 – Schrödinger's Cat Explained
13:12 – Can Quantum Effects Exist at a Macroscopic Scale?
17:45 – The Experiment That Changed Quantum Computing
34:33 – The Biggest Challenge: Scaling Quantum Computers
43:21 – John Martinis on Google's Quantum Supremacy
45:51 – AI vs Quantum Computing
01:01:45 – Can Quantum and Classical Computers Work Together?
01:07:36 – The NVIDIA Model for Quantum Computing
About TechSurge:
TechSurge Podcast shares the latest insights directly from legendary Silicon Valley leaders,
daring new founders, and visionary technologists.
Subscribe for weekly conversations into the intersection of technology advancement, market dynamics, and founder journeys.
#quantumcomputing #quantumphysics #nobelprize #technology

Silicon Valley was built on semiconductors, but for nearly two decades, venture capital shifted its attention towards software. Today, AI is changing that as the demand for compute, memory and networking explodes, hardware is once again at the centre of the industry's biggest bets. 

In this episode of TechSurge, host Michael Marks speaks with Lip-Bu Tan, CEO of Intel and one of the semiconductor industry's most influential investors and executives. The conversation traces Tan's journey from studying nuclear engineering at MIT to leading Cadence's turnaround, investing in more than 500 technology companies, and now steering Intel through one of the most significant transformations in its history.

Tan shares his VC conviction on backing semiconductor startups when most venture investors favored software, and why he believes AI's next breakthroughs will come from advances in memory, packaging, photonics, cooling and high-speed connectivity. He also opens up on the leadership philosophy that defined his time at Cadence, where listening to customers and building a culture of responsiveness became the foundation of the company's revival.

Wearing his CEO hat, Tan explains Intel's long-term strategy, why vertical integration still matters, how the company plans to reconnect with the startup ecosystem, and why missing another technology wave is not an option. 

Speaker Profiles and Links

Lip-Bu Tan: CEO of Intel Corporation, Chairman of Walden International, Founding Managing Partner of Walden Catalyst Ventures
LinkedIn: https://www.linkedin.com/in/lip-bu-tan-284a7846/
celesta.vc bio link 
Intel ceo bio link 

Further reading and resources
Reuters – “Intel’s new CEO plots overhaul of manufacturing and AI operations”https://www.reuters.com/technology/intels-new-ceo-plots-overhaul-manufacturing-ai-operations-2025-03-17/ 
Intel – https://www.intel.com
Celesta Capital – https://www.celesta.vc
SIA – “Global annual semiconductor sales increase 25.6% to $791.7 billion in 2025” – https://www.semiconductors.org/global-annual-semiconductor-sales-increase-25-6-to-791-7-billion-in-2025/
Infercom – “What is an RDU? Reconfigurable Dataflow Unit” – https://infercom.ai/glossary/rdu/
SemiconductorX – “Advanced Packaging: CoWoS, Foveros, EMIB, 3D IC” – https://semiconductorx.com/packaging-overview.html
TWIML AI Podcast – “Dataflow Computing for AI Inference [Kunle Olukotun]” – https://twimlai.com/go/751

Chapters:

00:00- Introduction
03:03- Lip-Bu Tan's Journey to Silicon Valley
04:12- Betting on Semiconductors Before AI
06:25- Why Hardware Matters Again
07:35- Investing in Deep Tech
09:31- Learning Through Boardrooms
12:10- Building the Next Generation of AI Infrastructure
17:03- The Cadence Turnaround
19:03- Customer Obsession as a Leadership Strategy
23:02- Rebuilding Intel
26:03- AI's Next Bottlenecks
30:32- Looking Ahead: The Future of Computing

Artificial intelligence is often discussed through models and GPUs. This episode looks beneath that surface, at the power delivery and networking required to make AI work at scale.
Host Sriram Viswanathan speaks with Rajiv Khemani, a serial deep tech entrepreneur whose career has tracked several major infrastructure cycles: internet networking, cloud switching, blockchain compute and now AI networking. Khemani reflects on his early work at NetBoost and Intel, his operating role at Cavium, and the founding of Innovium, which Marvell agreed to acquire for $1.1 billion in 2021. He also explains how work on low-power blockchain silicon led his team toward the infrastructure demands created by generative AI.
The discussion examines why incumbents often overlook emerging markets, why purpose-built hardware can outperform systems inherited from an earlier technology cycle, and how founders decide whether to keep financing a company or sell while the outcome remains attractive. Khemani describes the concentration risk of selling to a small number of hyperscalers, the fragility of semiconductor supply chains, and why leading-edge chip development now demands much larger balance sheets.
The conversation then turns to AI’s emerging bottlenecks. Large models require many accelerators to operate as one computer, making low-latency scale-up and scale-out networks central to performance. The episode explores heterogeneous compute, open networking standards, memory scarcity, AI’s growing electricity demand, and the competition between AI and Bitcoin mining for energy. 

Speaker Profiles and Links

Sriram Viswanathan: Founding Managing Partner, Celesta Capital — https://www.linkedin.com/in/onesriram/
Rajiv Khemani: Co-founder and Executive Chairman, Upscale AI; deep-tech entrepreneur and IIT Delhi alumnus
LinkedIn: https://www.linkedin.com/in/rajivkhemani/
Profile and contribution to the IIT, Delhi, Yardi School of Artificial Intelligence : https://scai.iitd.ac.in/rajiv-khemani 

References Mentioned and Further Reading

Upscale AI : https://upscaleai.com/ 
Upscale AI Launch Announcement : https://upscaleai.com/press-release/ 
Velaura AI : https://velaura.ai/ 
Acquisition of Innovium and cloud data-centre switching rationale, Marvell: https://www.marvell.com/company/newsroom/marvell-to-acquire-innovium-accelerates-cloud-growth-with-expanded-ethernet-switching-portfolio.html 
Cavium combination and infrastructure semiconductor strategy, Marvell:  https://www.marvell.com/company/newsroom/marvell-and-cavium-to-combine-creating-an-infrastructure-solutions-powerhouse.html 
Energy and AI, International Energy Agency: https://www.iea.org/reports/energy-and-ai
Energy demand from AI, International Energy Agency: https://www.iea.org/reports/energy-and-ai/energy-demand-from-ai 
Tokenisation in the context of money and other assets, Bank for International Settlements : https://www.bis.org/cpmi/publ/d225.pdf 
Leveraging tokenisation for payments and financial transactions, Bank for International Settlements : https://www.bis.org/publ/othp92.pdf 
Collective communication for clusters exceeding 100,000 GPUs, Meta researchers : https://arxiv.org/abs/2510.20171 
Load balancing for AI training workloads, UC Berkeley researchers : https://arxiv.org/abs/2507.21372 
Reliability in large-scale machine-learning clusters : https://arxiv.org/abs/2410.21680 
Bitcoin: A Peer-to-Peer Electronic Cash System : https://bitcoin.org/bitcoin.pdf 

Timestamps:

[Timestamp] Chapter Title
00:00 - Highlights and welcome
02:28 - From IIT Delhi to Silicon Valley
07:50 - Building Through Major Technology Waves
10:19 - Why Incumbents Miss Emerging Markets And Where Start-Ups Win
12:47 - Building Innovium for the Cloud
19:37 - Supply Shocks and Strategic Exits
27:28 - From Bitcoin Chips to AI
30:09 - Bitcoin, Tokenisation and Energy
42:37 - Agentic AI and Future Networks
53:21 - Memory, Capital and Founder Resilience