Jul 16, 2026 · 44m · mad

OpenAI’s Compute Chief: We Can’t Build Fast Enough | Sachin Katti

Sachin Katti · 27m spoken Matt Turck · 12m spoken
0:00 / 0:00
▶ Watch on YouTube →

gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions

In this episode of The MAD Podcast, OpenAI's Head of Compute Infrastructure Sachin Katti discusses the massive physical, financial, and technological engineering required to build and power gigawatt-scale AI data centers. He details OpenAI's multi-cloud strategy, custom silicon co-design, energy grid integration, and the physical constraints facing global AI deployment.

How this conversation actually went

Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Matt holds 29.9% of the talking time here. How this is scored →

Matt as informed peer 3.7 Guest teaching 5.0 Guest disagreement 1.8 Matt pushing back 1.7
05100:0015:0030:001:30–3:41 · Matt as informed peer 4/10 Historic Scale and Economics of AI Infrastructure Matt cites specific industry spend metrics like $50 billion for OpenAI and $700 billion across AI compute. Sachin confirms the scale of investments and notes the rapid pace of decision-making required inside OpenAI.3:41–8:10 · Matt as informed peer 3/10 OpenAI's Evolution Into Direct Infrastructure Construction Matt asks if OpenAI building datacenters represents a corporate pivot or new line of business. Sachin reframes this as developing a new organizational muscle, explaining the physical reality of liquid-cooled datacenters converting electrons to tokens.8:10–11:49 · Matt as informed peer 4/10 Power Grid Integration, On-Site Energy, and Nuclear Energy Matt connects their Paris location to France's nuclear infrastructure and asks about datacenter power sources. Sachin explains behind-the-meter gas generation and OpenAI's commitments to fund local grid generation and transmission.11:49–14:59 · Matt as informed peer 4/10 OpenAI's Custom Silicon and the Jalapeño Inference Chip Matt prompts Sachin on the strategy behind custom silicon like Jalapeno and whether inference compute now dominates training. Sachin educates Matt on tokens-per-watt efficiency and reframes how post-training synthetic generation blurs the line between training and inference.14:59–17:55 · Matt as informed peer 4/10 Overbuilding Risks vs. Physical Supply Chain Constraints Matt pushes on the potential risk of datacenter overbuilding due to long construction lead times. Sachin firmly counters that demand far outstrips supply and clarifies that his main paranoia is physical supply chain bottlenecks hindering expansion.17:55–21:15 · Matt as informed peer 4/10 Local Economic Impact and Closed-Loop Water Recycling Matt brings up public relations challenges regarding local community impact and water consumption myths. Sachin explains that closed-loop cooling systems consume minimal net water and highlights tax base and infrastructure benefits for rural areas.21:15–25:59 · Matt as informed peer 3/10 Site Selection Criteria for Next-Generation Data Centers Matt asks about datacenter site selection criteria and Sachin's role as Head of Industrial Compute. Sachin outlines key location criteria and describes the internal challenge of allocating scarce compute capacity among competing teams.25:59–28:05 · Matt as informed peer 2/10 Sachin Katti's Career Path Across Academia, Startups, and Industry Matt asks Sachin about his unique career path across academia, startups, and major tech firms. Sachin highlights his background as a Stanford professor and Intel CTO, noting how OpenAI uniquely merges research lab, startup, and scale.28:05–31:21 · Matt as informed peer 5/10 OpenAI's Infrastructure Partnerships and Project Stargate Matt demonstrates deep industry awareness by naming various compute partnerships and questioning if Project Stargate evolved. Sachin clarifies that Stargate serves as an overall umbrella strategy rather than a single joint venture.31:21–34:16 · Matt as informed peer 4/10 The Abilene Texas Supercluster and Expansion Sites Matt inquires about the Abilene supercluster and how datacenter buildouts are financed. Sachin explains that OpenAI operates as an off-take tenant while cloud and infrastructure partners handle the capital and construction.34:16–36:23 · Matt as informed peer 4/10 Accelerated Chip Development and AI-Assisted Hardware Design Matt notes the remarkably fast nine-month tapout timeline for the Jalapeno chip. Sachin attributes this speed to experienced talent, Broadcom partnership, knowing workload specifications, and using AI tools to accelerate hardware design.36:23–38:47 · Matt as informed peer 3/10 Interconnect Reliability and the Open-Source MRC Protocol Matt asks about the newly open-sourced MRC networking protocol. Sachin uses a city traffic routing analogy to explain how multi-path packet spraying ensures fault tolerance across massive 100,000 GPU clusters.38:47–42:08 · Matt as informed peer 4/10 Industrial Bottlenecks, Hardware Lead Times, and Labor Demands Matt brings up hardware bottlenecks like memory and electrician shortages, along with OpenAI's new guaranteed capacity offering. Sachin details long lead times for physical equipment and recontextualizes guaranteed capacity as securing enterprise AI supply.1:30–3:41 · Guest teaching 3/10 Historic Scale and Economics of AI Infrastructure Matt cites specific industry spend metrics like $50 billion for OpenAI and $700 billion across AI compute. Sachin confirms the scale of investments and notes the rapid pace of decision-making required inside OpenAI.3:41–8:10 · Guest teaching 5/10 OpenAI's Evolution Into Direct Infrastructure Construction Matt asks if OpenAI building datacenters represents a corporate pivot or new line of business. Sachin reframes this as developing a new organizational muscle, explaining the physical reality of liquid-cooled datacenters converting electrons to tokens.8:10–11:49 · Guest teaching 5/10 Power Grid Integration, On-Site Energy, and Nuclear Energy Matt connects their Paris location to France's nuclear infrastructure and asks about datacenter power sources. Sachin explains behind-the-meter gas generation and OpenAI's commitments to fund local grid generation and transmission.11:49–14:59 · Guest teaching 6/10 OpenAI's Custom Silicon and the Jalapeño Inference Chip Matt prompts Sachin on the strategy behind custom silicon like Jalapeno and whether inference compute now dominates training. Sachin educates Matt on tokens-per-watt efficiency and reframes how post-training synthetic generation blurs the line between training and inference.14:59–17:55 · Guest teaching 6/10 Overbuilding Risks vs. Physical Supply Chain Constraints Matt pushes on the potential risk of datacenter overbuilding due to long construction lead times. Sachin firmly counters that demand far outstrips supply and clarifies that his main paranoia is physical supply chain bottlenecks hindering expansion.17:55–21:15 · Guest teaching 5/10 Local Economic Impact and Closed-Loop Water Recycling Matt brings up public relations challenges regarding local community impact and water consumption myths. Sachin explains that closed-loop cooling systems consume minimal net water and highlights tax base and infrastructure benefits for rural areas.21:15–25:59 · Guest teaching 5/10 Site Selection Criteria for Next-Generation Data Centers Matt asks about datacenter site selection criteria and Sachin's role as Head of Industrial Compute. Sachin outlines key location criteria and describes the internal challenge of allocating scarce compute capacity among competing teams.25:59–28:05 · Guest teaching 3/10 Sachin Katti's Career Path Across Academia, Startups, and Industry Matt asks Sachin about his unique career path across academia, startups, and major tech firms. Sachin highlights his background as a Stanford professor and Intel CTO, noting how OpenAI uniquely merges research lab, startup, and scale.28:05–31:21 · Guest teaching 5/10 OpenAI's Infrastructure Partnerships and Project Stargate Matt demonstrates deep industry awareness by naming various compute partnerships and questioning if Project Stargate evolved. Sachin clarifies that Stargate serves as an overall umbrella strategy rather than a single joint venture.31:21–34:16 · Guest teaching 5/10 The Abilene Texas Supercluster and Expansion Sites Matt inquires about the Abilene supercluster and how datacenter buildouts are financed. Sachin explains that OpenAI operates as an off-take tenant while cloud and infrastructure partners handle the capital and construction.34:16–36:23 · Guest teaching 5/10 Accelerated Chip Development and AI-Assisted Hardware Design Matt notes the remarkably fast nine-month tapout timeline for the Jalapeno chip. Sachin attributes this speed to experienced talent, Broadcom partnership, knowing workload specifications, and using AI tools to accelerate hardware design.36:23–38:47 · Guest teaching 6/10 Interconnect Reliability and the Open-Source MRC Protocol Matt asks about the newly open-sourced MRC networking protocol. Sachin uses a city traffic routing analogy to explain how multi-path packet spraying ensures fault tolerance across massive 100,000 GPU clusters.38:47–42:08 · Guest teaching 6/10 Industrial Bottlenecks, Hardware Lead Times, and Labor Demands Matt brings up hardware bottlenecks like memory and electrician shortages, along with OpenAI's new guaranteed capacity offering. Sachin details long lead times for physical equipment and recontextualizes guaranteed capacity as securing enterprise AI supply.1:30–3:41 · Guest disagreement 1/10 Historic Scale and Economics of AI Infrastructure Matt cites specific industry spend metrics like $50 billion for OpenAI and $700 billion across AI compute. Sachin confirms the scale of investments and notes the rapid pace of decision-making required inside OpenAI.3:41–8:10 · Guest disagreement 2/10 OpenAI's Evolution Into Direct Infrastructure Construction Matt asks if OpenAI building datacenters represents a corporate pivot or new line of business. Sachin reframes this as developing a new organizational muscle, explaining the physical reality of liquid-cooled datacenters converting electrons to tokens.8:10–11:49 · Guest disagreement 1/10 Power Grid Integration, On-Site Energy, and Nuclear Energy Matt connects their Paris location to France's nuclear infrastructure and asks about datacenter power sources. Sachin explains behind-the-meter gas generation and OpenAI's commitments to fund local grid generation and transmission.11:49–14:59 · Guest disagreement 2/10 OpenAI's Custom Silicon and the Jalapeño Inference Chip Matt prompts Sachin on the strategy behind custom silicon like Jalapeno and whether inference compute now dominates training. Sachin educates Matt on tokens-per-watt efficiency and reframes how post-training synthetic generation blurs the line between training and inference.14:59–17:55 · Guest disagreement 4/10 Overbuilding Risks vs. Physical Supply Chain Constraints Matt pushes on the potential risk of datacenter overbuilding due to long construction lead times. Sachin firmly counters that demand far outstrips supply and clarifies that his main paranoia is physical supply chain bottlenecks hindering expansion.17:55–21:15 · Guest disagreement 2/10 Local Economic Impact and Closed-Loop Water Recycling Matt brings up public relations challenges regarding local community impact and water consumption myths. Sachin explains that closed-loop cooling systems consume minimal net water and highlights tax base and infrastructure benefits for rural areas.21:15–25:59 · Guest disagreement 1/10 Site Selection Criteria for Next-Generation Data Centers Matt asks about datacenter site selection criteria and Sachin's role as Head of Industrial Compute. Sachin outlines key location criteria and describes the internal challenge of allocating scarce compute capacity among competing teams.25:59–28:05 · Guest disagreement 1/10 Sachin Katti's Career Path Across Academia, Startups, and Industry Matt asks Sachin about his unique career path across academia, startups, and major tech firms. Sachin highlights his background as a Stanford professor and Intel CTO, noting how OpenAI uniquely merges research lab, startup, and scale.28:05–31:21 · Guest disagreement 3/10 OpenAI's Infrastructure Partnerships and Project Stargate Matt demonstrates deep industry awareness by naming various compute partnerships and questioning if Project Stargate evolved. Sachin clarifies that Stargate serves as an overall umbrella strategy rather than a single joint venture.31:21–34:16 · Guest disagreement 2/10 The Abilene Texas Supercluster and Expansion Sites Matt inquires about the Abilene supercluster and how datacenter buildouts are financed. Sachin explains that OpenAI operates as an off-take tenant while cloud and infrastructure partners handle the capital and construction.34:16–36:23 · Guest disagreement 1/10 Accelerated Chip Development and AI-Assisted Hardware Design Matt notes the remarkably fast nine-month tapout timeline for the Jalapeno chip. Sachin attributes this speed to experienced talent, Broadcom partnership, knowing workload specifications, and using AI tools to accelerate hardware design.36:23–38:47 · Guest disagreement 1/10 Interconnect Reliability and the Open-Source MRC Protocol Matt asks about the newly open-sourced MRC networking protocol. Sachin uses a city traffic routing analogy to explain how multi-path packet spraying ensures fault tolerance across massive 100,000 GPU clusters.38:47–42:08 · Guest disagreement 2/10 Industrial Bottlenecks, Hardware Lead Times, and Labor Demands Matt brings up hardware bottlenecks like memory and electrician shortages, along with OpenAI's new guaranteed capacity offering. Sachin details long lead times for physical equipment and recontextualizes guaranteed capacity as securing enterprise AI supply.1:30–3:41 · Matt pushing back 1/10 Historic Scale and Economics of AI Infrastructure Matt cites specific industry spend metrics like $50 billion for OpenAI and $700 billion across AI compute. Sachin confirms the scale of investments and notes the rapid pace of decision-making required inside OpenAI.3:41–8:10 · Matt pushing back 2/10 OpenAI's Evolution Into Direct Infrastructure Construction Matt asks if OpenAI building datacenters represents a corporate pivot or new line of business. Sachin reframes this as developing a new organizational muscle, explaining the physical reality of liquid-cooled datacenters converting electrons to tokens.8:10–11:49 · Matt pushing back 1/10 Power Grid Integration, On-Site Energy, and Nuclear Energy Matt connects their Paris location to France's nuclear infrastructure and asks about datacenter power sources. Sachin explains behind-the-meter gas generation and OpenAI's commitments to fund local grid generation and transmission.11:49–14:59 · Matt pushing back 2/10 OpenAI's Custom Silicon and the Jalapeño Inference Chip Matt prompts Sachin on the strategy behind custom silicon like Jalapeno and whether inference compute now dominates training. Sachin educates Matt on tokens-per-watt efficiency and reframes how post-training synthetic generation blurs the line between training and inference.14:59–17:55 · Matt pushing back 3/10 Overbuilding Risks vs. Physical Supply Chain Constraints Matt pushes on the potential risk of datacenter overbuilding due to long construction lead times. Sachin firmly counters that demand far outstrips supply and clarifies that his main paranoia is physical supply chain bottlenecks hindering expansion.17:55–21:15 · Matt pushing back 2/10 Local Economic Impact and Closed-Loop Water Recycling Matt brings up public relations challenges regarding local community impact and water consumption myths. Sachin explains that closed-loop cooling systems consume minimal net water and highlights tax base and infrastructure benefits for rural areas.21:15–25:59 · Matt pushing back 1/10 Site Selection Criteria for Next-Generation Data Centers Matt asks about datacenter site selection criteria and Sachin's role as Head of Industrial Compute. Sachin outlines key location criteria and describes the internal challenge of allocating scarce compute capacity among competing teams.25:59–28:05 · Matt pushing back 1/10 Sachin Katti's Career Path Across Academia, Startups, and Industry Matt asks Sachin about his unique career path across academia, startups, and major tech firms. Sachin highlights his background as a Stanford professor and Intel CTO, noting how OpenAI uniquely merges research lab, startup, and scale.28:05–31:21 · Matt pushing back 3/10 OpenAI's Infrastructure Partnerships and Project Stargate Matt demonstrates deep industry awareness by naming various compute partnerships and questioning if Project Stargate evolved. Sachin clarifies that Stargate serves as an overall umbrella strategy rather than a single joint venture.31:21–34:16 · Matt pushing back 2/10 The Abilene Texas Supercluster and Expansion Sites Matt inquires about the Abilene supercluster and how datacenter buildouts are financed. Sachin explains that OpenAI operates as an off-take tenant while cloud and infrastructure partners handle the capital and construction.34:16–36:23 · Matt pushing back 1/10 Accelerated Chip Development and AI-Assisted Hardware Design Matt notes the remarkably fast nine-month tapout timeline for the Jalapeno chip. Sachin attributes this speed to experienced talent, Broadcom partnership, knowing workload specifications, and using AI tools to accelerate hardware design.36:23–38:47 · Matt pushing back 1/10 Interconnect Reliability and the Open-Source MRC Protocol Matt asks about the newly open-sourced MRC networking protocol. Sachin uses a city traffic routing analogy to explain how multi-path packet spraying ensures fault tolerance across massive 100,000 GPU clusters.38:47–42:08 · Matt pushing back 2/10 Industrial Bottlenecks, Hardware Lead Times, and Labor Demands Matt brings up hardware bottlenecks like memory and electrician shortages, along with OpenAI's new guaranteed capacity offering. Sachin details long lead times for physical equipment and recontextualizes guaranteed capacity as securing enterprise AI supply.

speaking balance: gold is Matt, purple is the guest (3 minute bins)

0:00 · Matt 55.7% · guest 44.3%0:00 · Matt 55.7% · guest 44.3%3:00 · Matt 45.6% · guest 54.4%3:00 · Matt 45.6% · guest 54.4%6:00 · Matt 21.2% · guest 78.8%6:00 · Matt 21.2% · guest 78.8%9:00 · Matt 17.7% · guest 82.3%9:00 · Matt 17.7% · guest 82.3%12:00 · Matt 36.3% · guest 63.7%12:00 · Matt 36.3% · guest 63.7%15:00 · Matt 28.2% · guest 71.8%15:00 · Matt 28.2% · guest 71.8%18:00 · Matt 23.3% · guest 76.7%18:00 · Matt 23.3% · guest 76.7%21:00 · Matt 33.6% · guest 66.4%21:00 · Matt 33.6% · guest 66.4%24:00 · Matt 19.5% · guest 80.5%24:00 · Matt 19.5% · guest 80.5%27:00 · Matt 31.6% · guest 68.4%27:00 · Matt 31.6% · guest 68.4%30:00 · Matt 20.5% · guest 79.5%30:00 · Matt 20.5% · guest 79.5%33:00 · Matt 39.5% · guest 60.5%33:00 · Matt 39.5% · guest 60.5%36:00 · Matt 16.6% · guest 83.4%36:00 · Matt 16.6% · guest 83.4%39:00 · Matt 23.8% · guest 76.2%39:00 · Matt 23.8% · guest 76.2%42:00 · Matt 37.7% · guest 62.3%42:00 · Matt 37.7% · guest 62.3%
Sharpest disagreement ▶ 17:17 Rejecting overbuilding concerns

Sachin directly rejects the host's framing of potential overbuilding risk, clarifying that demand far outstrips supply and that physical supply chain delays are his actual operational paranoia.

Hardest push from Matt ▶ 14:59 Challenging overbuilding and demand lag

Matt directly challenges Sachin on the risk of overbuilding compute infrastructure, pressing on the inherent delay between constructing physical datacenters and actual demand realization.

Biggest teaching moment ▶ 14:16 Educating on inference versus training

Sachin reframes the host's clear-cut separation between training and inference workloads, explaining that synthetic data generation, post-training, and test-time compute mean much of modern training is fundamentally inference.

Matt holds his own ▶ 28:05 Detailing multi-billion dollar compute ecosystem

Matt demonstrates notable expertise by citing specific deal figures, including a $20 billion arrangement with Cerebras, and pressing Sachin on how Stargate fits into OpenAI's broader partner landscape.

the scores for every segment, with the reasoning behind each
ChapterTopicMatt as informed peerGuest teachingGuest disagreementMatt pushing backWhy
Historic Scale and Economics of AI Infrastructure 4311 Matt cites specific industry spend metrics like $50 billion for OpenAI and $700 billion across AI compute. Sachin confirms the scale of investments and notes the rapid pace of decision-making required inside OpenAI.
OpenAI's Evolution Into Direct Infrastructure Construction 3522 Matt asks if OpenAI building datacenters represents a corporate pivot or new line of business. Sachin reframes this as developing a new organizational muscle, explaining the physical reality of liquid-cooled datacenters converting electrons to tokens.
Power Grid Integration, On-Site Energy, and Nuclear Energy 4511 Matt connects their Paris location to France's nuclear infrastructure and asks about datacenter power sources. Sachin explains behind-the-meter gas generation and OpenAI's commitments to fund local grid generation and transmission.
OpenAI's Custom Silicon and the Jalapeño Inference Chip 4622 Matt prompts Sachin on the strategy behind custom silicon like Jalapeno and whether inference compute now dominates training. Sachin educates Matt on tokens-per-watt efficiency and reframes how post-training synthetic generation blurs the line between training and inference.
Overbuilding Risks vs. Physical Supply Chain Constraints 4643 Matt pushes on the potential risk of datacenter overbuilding due to long construction lead times. Sachin firmly counters that demand far outstrips supply and clarifies that his main paranoia is physical supply chain bottlenecks hindering expansion.
Local Economic Impact and Closed-Loop Water Recycling 4522 Matt brings up public relations challenges regarding local community impact and water consumption myths. Sachin explains that closed-loop cooling systems consume minimal net water and highlights tax base and infrastructure benefits for rural areas.
Site Selection Criteria for Next-Generation Data Centers 3511 Matt asks about datacenter site selection criteria and Sachin's role as Head of Industrial Compute. Sachin outlines key location criteria and describes the internal challenge of allocating scarce compute capacity among competing teams.
Sachin Katti's Career Path Across Academia, Startups, and Industry 2311 Matt asks Sachin about his unique career path across academia, startups, and major tech firms. Sachin highlights his background as a Stanford professor and Intel CTO, noting how OpenAI uniquely merges research lab, startup, and scale.
OpenAI's Infrastructure Partnerships and Project Stargate 5533 Matt demonstrates deep industry awareness by naming various compute partnerships and questioning if Project Stargate evolved. Sachin clarifies that Stargate serves as an overall umbrella strategy rather than a single joint venture.
The Abilene Texas Supercluster and Expansion Sites 4522 Matt inquires about the Abilene supercluster and how datacenter buildouts are financed. Sachin explains that OpenAI operates as an off-take tenant while cloud and infrastructure partners handle the capital and construction.
Accelerated Chip Development and AI-Assisted Hardware Design 4511 Matt notes the remarkably fast nine-month tapout timeline for the Jalapeno chip. Sachin attributes this speed to experienced talent, Broadcom partnership, knowing workload specifications, and using AI tools to accelerate hardware design.
Interconnect Reliability and the Open-Source MRC Protocol 3611 Matt asks about the newly open-sourced MRC networking protocol. Sachin uses a city traffic routing analogy to explain how multi-path packet spraying ensures fault tolerance across massive 100,000 GPU clusters.
Industrial Bottlenecks, Hardware Lead Times, and Labor Demands 4622 Matt brings up hardware bottlenecks like memory and electrician shortages, along with OpenAI's new guaranteed capacity offering. Sachin details long lead times for physical equipment and recontextualizes guaranteed capacity as securing enterprise AI supply.

Statements from this episode (21)

Opinion
Katti: AI compute buildout is among largest human projects in history
“It definitely feels like one of the largest things humanity has ever built, effectively.”
Sachin Katti Jul 16, 2026 ▶ 2:12
Disclosure
Katti: OpenAI on track to spend roughly $50B on compute this year
“Actually, that sounds okay.”
Sachin Katti Jul 16, 2026 ▶ 3:15
Disclosure
Katti: OpenAI must build its own compute infrastructure alongside partners
“I think what's becoming clear is to build the kind of compute we need and at this scale we have to not just rely on getting compute from our partners. We increasingly have to take a much more active role in building and getting that compute that we need.”
Sachin Katti Jul 16, 2026 ▶ 4:30
Disclosure
Katti: OpenAI commits to not taking existing power from local grids
“Whenever we build a data center anywhere, we make it a hard commitment that we are not taking power away from the grid. In fact, we are investing in the grid to generate new power so that we can consume it for data centers.”
Sachin Katti Jul 16, 2026 ▶ 8:45
Opinion
Katti: Nuclear power is the densest, cleanest energy for AI data centers
“I think it is the densest form of energy we can all produce and consume and it's also clean. So I think definitely would be a good source of massive scalable energy for our data centers.”
Sachin Katti Jul 16, 2026 ▶ 11:20
Assertion Not checkable as stated
OpenAI's Jalapeño chip optimizes tokens produced per watt
“So the key metric that Jalapeno is optimizing is maximizing the number of tokens you can produce per watt.”
Sachin Katti Jul 16, 2026 ▶ 13:19
Assertion Not checkable as stated
Inference likely accounts for the majority of OpenAI's total compute
“Inference is a big, perhaps even the majority on compute.”
Sachin Katti Jul 16, 2026 ▶ 14:16
Insight
Katti: Modern AI model training consists heavily of inference workloads
“We don't like to make a distinction between Training and infants, because a lot of training is now infants. So when we train a new model, we are generating synthetic data, for example. That's inference. When we train a new model, we are doing post-train, and t…”
Sachin Katti Jul 16, 2026 ▶ 14:22
Assertion Not checkable as stated
Sachin Katti: OpenAI tripled compute and tripled revenue
“We tripled compute and we tripled revenue.”
Sachin Katti Jul 16, 2026 ▶ 15:53
Assertion Not checkable as stated
Katti: Demand far outstrips OpenAI's compute supply, zero goes to waste
“Demand far outstrips. Compute supply today. So anything we can bring online, we consume immediately. So there's no compute that is going to waste for us.”
Sachin Katti Jul 16, 2026 ▶ 15:56
Assertion Partly supported
Katti: Data centers do not net consume new water due to recycling
“It's a misperception that data centers consume a lot of water. It's anything. They consume so little water for what they do. And all of that water is recycled. So we don't net consume new water.”
Sachin Katti Jul 16, 2026 ▶ 20:46
Assertion Supported
Katti: OpenAI sources compute from all major cloud hyperscalers
“We have compute from, effectively, many sources. So, Microsoft, obviously, is a big partner, important partner. We also have compute, as we have announced, from AWS, Ion Google. So, we have compute from all of the hyperscalers, effectively.”
Sachin Katti Jul 16, 2026 ▶ 28:35
Disclosure
Katti: OpenAI considers designing and building its own data centers
“As we go forward obviously there'll be building on all of these relationships but also looking at more options where we design the compute, the data center itself, also potentially even build the data center ourselves.”
Sachin Katti Jul 16, 2026 ▶ 29:08
Assertion Supported
Katti: OpenAI's Abilene data center is operational and training latest models
“It's up and running. It's being used for training the last two models more actually.”
Sachin Katti Jul 16, 2026 ▶ 31:56
Assertion Supported
OpenAI taped out its custom Jalapeño chip in just nine months
“Yes, it was incredibly quick. Nine months is very, very fast.”
Sachin Katti Jul 16, 2026 ▶ 34:36
Assertion Supported
Katti: Many OpenAI chip team members previously designed Google TPUs
“They have, many, many of the team have designed TPU chips at Google in the past.”
Sachin Katti Jul 16, 2026 ▶ 34:51
Prediction Not checkable as stated
Katti: AI will soon design systems and chips for next-gen AI
“We do believe that the world of Likersh is not that far, where AI will design the systems it needs to train and run the next generation of AI.”
Sachin Katti Jul 16, 2026 ▶ 36:07
Assertion Supported
Katti: Turbine and transformer manufacturers face multi-year lead times to expand capacity
“Those industries have historically have not added much capacity for the last decade or so ago, and they've suddenly experienced a demand shock. And it takes years before you can add capacity to produce more turbines and transformers.”
Sachin Katti Jul 16, 2026 ▶ 39:18
Prediction Held up
Katti: Skilled labor shortages will become an increasing bottleneck for AI expansion
“No, I think there is definitely a shortage of electricians, plumbers, all kinds of trades, you name it. So anything we can do to train more folks to be able to do those, there are very well-paying jobs that a lot of us, all of the hyperscalers, all of the labs…”
Sachin Katti Jul 16, 2026 ▶ 39:55
Prediction Not checkable as stated
Sachin Katti: AI tokens will always command a premium due to compute shortages
“So in a world where computers are shortage, therefore, tokens are always going to be at a premium, and there's a shortage of tokens that we can produce, given the limited compute that we have”
Sachin Katti Jul 16, 2026 ▶ 41:07
Prediction Not checkable as stated
Orbital compute will become a component in the AI infrastructure arsenal
“Whether it is needed I think there is room for orbital compute. I don't think it's going to serve all the compute beats, but it's definitely going to be a component in the arsenal.”
Sachin Katti Jul 16, 2026 ▶ 42:52
Made with StarZero

Turn any episode into a week of clips.

This entire site, over 400 conversations transcribed, diarized, checked and made playable, runs on the StarZero media pipeline. Drop in your own episode and the podcast clipper finds the moments worth sharing, cuts them, captions them, and reframes them for every feed.