Wei Xie (@WeiInvests) header
Wei Xie (@WeiInvests) profile picture

Institutional investor. Ex-founder (AI agents, robotics infra). Notes from both chairs on investing in AI, tokenization, longevity. Views my own. Not advice.

Joined September 2021
Photo by Wei Xie (@WeiInvests)Photo by Wei Xie (@WeiInvests)Photo by Wei Xie (@WeiInvests)Photo by Wei Xie (@WeiInvests)Photo by Wei Xie (@WeiInvests)Photo by Wei Xie (@WeiInvests)Photo by Wei Xie (@WeiInvests)Photo by Wei Xie (@WeiInvests)Photo by Wei Xie (@WeiInvests)Photo by Wei Xie (@WeiInvests)
An $18.7Bn Market Failure: The Most Valuable Unclaimed Category in Robotics. The global robotics market is a multi-trillion dollar opportunity, yet progress is being stifled by a massive, obvious gap: the difference between marketing hype and actual autonomy. The industry is currently stuck in the "Demo-ware' era. This isn't just an engineering problem… it's a form of market failure. Robotics lacks the critical, independent evaluation infrastructure that accelerated the LLM revolution. This failure represents the most valuable unclaimed category in embodied AI. Let's break down the problem, the blueprint, and the staggering commercial opportunity that awaits.
12
7
1
43
7,843
nvidia's $36B "AI cloud agreements" might be the most interesting contract in the 10-Q, and almost nobody reads it like a lender would how it works, straight from the July-quarter filing: - an AI cloud buys nvidia systems - nvidia commits to rent that capacity back, deals typically about 6 years - the cloud can stop providing it to nvidia whenever it wants and sell to someone else at a better rate - nvidia's commitment shrinks as outside customers use the capacity, and nvidia can earn a revenue share now sit in the seat of whoever lent to that cloud. demand strong, the cloud sells to customers at higher prices. demand weak, nvidia is still on the hook to pay. that's a floor under the cloud's revenue, written by the company that sold it the chips so a chunk of neocloud credit quality starts to look like nvidia credit quality. a lender can underwrite that floor, not just the GPUs everyone's pricing "jensen out = less nvidia financing." fine. but the $36B already signed doesn't vanish with him. the question is whether new ones get written. if they slow, the next neocloud loan gets priced without that floor, on the GPUs and the cloud's own customers. that's a higher rate today's print matters, but the line I'd check is that $36B when the next 10-Q lands. it grows with new deals and shrinks as outside customers soak up the capacity, so the number alone won't tell you. listen for which one management credits
224
nvidia reports tuesday. the line I'm watching isn't revenue it's the support stack in the 10-Q: $99B stakes in its ecosystem up to $108.5B guarantees (105 = OpenAI Ohio, live 2028+) $36B AI-cloud capacity it backstops $25B investments promised, unpaid maybe some clues in how they present the info.
3
344
Markets underprice how resourceful people get when cornered. It's a game: one side plays a hand, the other finds the counter. CoreWeave's debt is stretched, so customers now prepay for capacity, some for the full contract upfront. That's financing. Management said it themselves: prepayments now beat the debt markets. Same with stretched balance sheets. The ratios scream default. But lenders want loans performing, so they amend and extend, especially in a downturn. Incentives beat ratios. The catch: risk doesn't disappear. It moves. The weak point now sits with the customers prepaying for compute. Equity or debt funded? Leverage in the system hasn't dropped. It continues to move.
1
468
Odd loop this week. Seeing reports on X that Nvidia is putting $10B into the Anthropic deal. If some of the Nvidia selling is people raising cash for the IPO, and most of what Anthropic raises goes back out as compute: Sell the chipmaker, fund the customer, customer buys chips.
1
2
449
This week we're talking vision and robots.
SCORE (SN44) is now a top 5 subnet on TAO There's not many people I admire more that Max to discuss the market & their hardwork Locked in, focused on taking Bittensor to the masses Reading FC AVIA PwC France As always, Full ep on Youtube 00:00 Market Discussion 01:08 Web2 markets 02:51 SCORE (SN44) entering $120b chocolate market 12:52 History and Evolution of SCORE (SN44) 17:33 Robotics: The Final Boss of AGI 27:45 The New Age of Compute 33:38 Visual Learning in Robotics 38:24 Closing
4
1
5
1,013
Wei Xie retweeted
SCORE (SN44) is now a top 5 subnet on TAO There's not many people I admire more that Max to discuss the market & their hardwork Locked in, focused on taking Bittensor to the masses Reading FC AVIA PwC France As always, Full ep on Youtube 00:00 Market Discussion 01:08 Web2 markets 02:51 SCORE (SN44) entering $120b chocolate market 12:52 History and Evolution of SCORE (SN44) 17:33 Robotics: The Final Boss of AGI 27:45 The New Age of Compute 33:38 Visual Learning in Robotics 38:24 Closing
6
9
4
65
7,686
Wei Xie retweeted
We're back on Supercycle discussing market trends AI stocks continue to rip, $TAO and $TIG becoming leading AI names and $CHZ ushering next gen ownership for sports Only on the Supercycle
3
7
4
40
5,789
Wei Xie retweeted
I love speaking with @here4impact, who spent his career on Wall St prior to investing his fund in bittensor:native Michael's #1 rule? Cashflow is King. "At the end of the day, all that matters is cash flow... Eventually if you don't generate cash, you are going out of business. That's just the way it works. There's no way around it."
Impressed by SN8 Vanta @0xarrash on their suite of trading products: "Our largest competitor, FTMO, makes $500mm roughly in registration revenue/year, and they're making a ton of money, but traders are being kind of taken advantage of We want to build an alternative that actually rewards traders."
2
4
39
3,475
Wei Xie retweeted
Score (SN44) Founder @MaxScore shares his commitment to Bittensor: "The switching costs aren't just technical, they're existential. Find the right incentive mechanism, get it running at the right pace, and you never touch it again. Right commodity, right price, right speed, right accuracy. You don't walk away from that."
3
8
78
3,117
Wei Xie retweeted
After the Templar Covenant exit last week, the need for institutional grade risk frameworks couldn’t be more obvious Since then we’ve seen Manako partner with PWC and multiple plans to rollout trillion parameter training runs from other subnets Compute is being squeezed and non-AI companies are pivoting to GPU marketplaces to capture this insatiable demand (see All Birds Inc yesterday) Capital is flowing into Bittensor subnets with minimal structure in evaluating signal. That’s how you get blindsided @TAOInstitute_ gives institutional allocators the risk frameworks and evaluation tooling to actually underwrite these positions Think of it as the infrastructure layer between capital and the subnet economy We are now live; so kick the tyres on the platform and tell us what’s missing
This quoted post is unavailable.
20
16
1
143
18,230
TAO Institute brings independent research, real-time risk analytics, and subnet-level coverage to the Bittensor ecosystem. Institutional-grade infrastructure for the open source AI economy.
Today, we launch TAO Institute Day of celebration Also sat down with Max from SCORE (SN44) Walking through the contract Manako has with PwC France and this is amazing Direct distribution to their clients, already with many clients under discussions for Manako Never seen anything like it before from a big 4 Powered by $TAO Participating in the open source economy Verified first on @TAOInstitute_
1
5
691
Wei Xie retweeted
TAO Institute is live and the first big milestone they covered is to verify our latest announcement. Long Live @TAOInstitute_
Today, we launch TAO Institute Day of celebration Also sat down with Max from SCORE (SN44) Walking through the contract Manako has with PwC France and this is amazing Direct distribution to their clients, already with many clients under discussions for Manako Never seen anything like it before from a big 4 Powered by $TAO Participating in the open source economy Verified first on @TAOInstitute_
3
8
82
3,916
Wei Xie retweeted
Today, we launch TAO Institute Day of celebration Also sat down with Max from SCORE (SN44) Walking through the contract Manako has with PwC France and this is amazing Direct distribution to their clients, already with many clients under discussions for Manako Never seen anything like it before from a big 4 Powered by $TAO Participating in the open source economy Verified first on @TAOInstitute_
This quoted post is unavailable.
19
29
7
229
26,508
Why did Templar (SN3) leave $TAO? Is this the end for Bittensor? When emerging technologies grow, it is not often linear; There's evolution, mishaps and constant iterations. Such is life. We spoke with Max from Score (SN44) and Gyles Founder of Tensora Group to share their thoughts, two of the most compelling founders in Bittensor. More eps with TAO investors and founders to share their outlook on the future, follow @supercyclepod to get it first. Timestamps 00:00 Introductions 04:27 Initial reactions to Templar SN3 07:11 Why Bittensor? 11:29 History of Bittensor 18:50 Bittensor's path to Hyperscale 25:16 Bittensor's endless ecosystem developments 28:48 Mission of Decentralized AI 32:40 Closing Thoughts
4
11
3
70
11,568
Wei Xie retweeted
Metanova (SN68) is one of the most compelling desci tokens in crypto, uniquely enabled by $TAO.
I only invest in $TAO now It is the most compelling ecosystem in crypto with the highest rate of product iteration and talent density I've ever come across My discussion with Wei below on @supercyclepod 00:00 Introduction 01:50 Macro Outlook 09:48 Speculative demand shift from crypto to AI 12:31 JP Morgan and the institutionalization of blockchain 15:15 AI vs. crypto: intelligence amplifiers vs. value representation 22:50 Korea's "Salaryman" dream and social drivers of crypto volume 33:00 Discovering Bittensor and decentralized systems 41:42 Why decentralized AI matters as a counterweight to frontier labs 47:53 Metanova (SN68) and AI drug discovery on Bittensor 01:04:32 Babelbit (SN59), Yanez (SN54), Beam (SN105) 01:08:45 Why founders choose Bittensor over traditional VC
2
23
2
168
16,708
Robotics is a trillion dollar opportunity, but most people can't participate Tens of billions of value has already been created in private markets Figure AI went from $2B to $39B in just one year Crypto likely has a unique oppurtunity to provide access to millions of people into the new robotic revolution via @xmaquina Mauricio from XMAQUINA and Simon from Moonrock Capital joined us on the pod to discuss the current landscape 00:12 Intro 01:24 Origin Story 07:51 Bull Case for Robotics 11:11 Macroeconomic Outlook 16:16 Intersection of Crypto and Robotics 19:21 XMAQ Portfolio Deep Dive: Apptroniks, Agility, 1x 41:52 Future of XMAQ 56:19 Long Live Token Launches 01:06:13 Rapid Fire Round
5
15
4
88
28,391
TL;DR Photometric robustness in VLA models is achievable. One model already proved it. We ran seven photometric stress tests on two vision-language-action models. Same benchmark, same perturbations, same severity levels. Pi 0.5 held flat. SmolVLA lost ground on nearly every one.
3
3
10
827
Why that matters? A model that works in the lab and breaks when the lighting shifts WILL break when deployed under a different sky. Full write-up and interactive perturbation explorer can be found on our site: nitter.cf/sihing_guppy/status/20…
2
2
7
351