Building New Financial Markets for the AI Value Chain.
USA
Joined December 2011
- Tweets5.5K
- Following379
- Followers767
- Likes13.9K
Ash | Carry retweeted
Be the Seminole Tribe of Florida.
> Build a bingo hall on your Hollywood reservation for approximately $900,000 in 1979.
> Expand into casinos. Open Hard Rock hotel and casino resorts in Tampa and Hollywood in 2004.
> In 2007, acquire Hard Rock International for $965 million.
> The business includes 124 cafés, hotels, casinos, entertainment venues and a globally recognized brand.
> In 28 years, go from operating a bingo hall to acquiring an international hospitality company.
> Continue expanding. Hard Rock now has venues in more than 70 countries.
> In 2022, pay another $1.075 billion for the operating assets of the Mirage in Las Vegas, with the real estate leased separately.
> Keep the parent company under tribal ownership.
> Use gaming revenue to fund more than 90% of the tribal government’s budget, supporting healthcare, education, housing, public safety and other services.
> Start with a local business. Expand operations. Acquire the brand. Build internationally. Retain ownership and use the income to support your community.
One of the more remarkable business success stories in America!
Today I issued an Executive Order designating “Seminole Wind” by John Anderson as the official anthem of Florida.
I have also directed the @FLSecofState to erect historic markers in the City of Apopka depicting John Anderson’s contribution to Florida’s heritage.
In addition, I have tasked @FLDEPNews to work with the @SJRWMD to designate wildlife drive and trail system within Lake Apopka North Shore as the John Anderson Wildlife Drive and Trail System.
This is the numero uno indicator of where inference market is going.
The number of customers paying >100K to Anthropic went up from 20 to 12,000 in less than 2 years.
What happens when this number shoots to 12,000,000.
Ash | Carry retweeted
Unlike the dot-com bubble, when global technology stocks surged far ahead of earnings, the current rally has been supported by stronger earnings growth.
Source: @steve_donze
Why we so often end up arguing with LLMs despite knowing that it can't really think independently?
Or is it just me?
Ash | Carry retweeted
Good data showing that NVIDIA AI infrastructure is a durable, investable asset that retains value.
New applications and workloads keep landing on our fungible platform, increasing demand because of its productivity.
Chart by @a16z
GPUs are aging like fine wine, with rental rates up across the board since January 1st
Approximate change YTD:
- H200: +160%
- B200: +120%
- RTX PRO 6000: +65%
- H100: +35%
Even the 6-year-old A100 rents for more than it did at the start of the year
Charts of the Week: a16z.news/p/charts-of-the-we…
Ash | Carry retweeted
15 business leaders on AI infrastructure demand and capacity in the last ~month or so
>@ajassy: “2027 is largely sold out. … A good chunk of 2028 is sold out.”
>@sundarpichai: “We are capacity constrained on a daily basis”
>@JensenHuang: “Demand is extraordinary. Capacity is the constraint.” On future optics needs: “several orders of magnitude more than we currently ship today”
>Matt Garman, AWS: “the demand is just massive”
>C.C. Wei, TSMC: “we did not see any sign of slowdown”
>@LisaSu: “There’s very, very high demand for the next several years”
>Sanjay Mehrotra, Micron, on DRAM: “we do not have line of sight to when supply and demand will return to balance”
>Christophe Fouquet, ASML: Planned 2027–28 supply “will not fully meet the demand of our customers”
>Michael Hurlston, Lumentum: “completely sold out”. Bloomberg reports the optical-component horizon extends through early 2029
>@finkd: “we have not yet found the limit” of applying compute to Meta’s core business
>Josh Payne, Nscale: “unprecedented global demand”
>Tim Archer, Lam Research: “AI driving higher semiconductor demand and greater device complexity”
>@satyanadella: “I am confident in the long-term returns on our investments.” Based on demand and efficiency gains
>@LipBuTan1: “We don’t see any slowdown”
>Hock Tan, Broadcom: “The thing you hit on that I think might be the biggest constraint is capital”
Memory ETFs not only lost the bullish momentum but are also facing outflow pressures now as prices keep moving lower.
From $40 to $81 in 3 months and back to $47 in next 3. What a ride!
ALT Memory stock ETF $DRAM performance and inflow(outflow)
Sentience or human brain like performance may not be really coming anytime soon, despite AI industry's warning calls.
A chart of efficiency and performance divergence between biological systems and artificial models as scale increases.
Ash | Carry retweeted
AI is generating massive fortunes in the US and China:
The combined wealth of new US AI billionaire entrants rose +$44 billion in the first 9 months of 2026, to $132 billion.
Over the same time period, the combined wealth of new Chinese AI billionaire entrants surged +$43 billion, to $89 billion.
This metric tracks the combined net worth of people who newly entered the world’s 500 richest list in 2026 through AI-related businesses.
This includes 12 billionaires from the US and 5 from China.
As a result, the combined wealth of this group in the US and China soared +$87 billion, to $221 billion.
AI is creating unprecedented amounts of wealth.
Ash | Carry retweeted
Trafigura says hyperscaler plans imply AI copper demand of 2.5-3 Mt in 2028, up from about 1 Mt this year
Henry Van, Trafigura's head of industrial metals analysis, called that the full-build case. Market pricing implies 1.6 Mt, still up 60%.
Chief economist Saad Rahim said global copper demand would fall this year without data centers. Van said data center builders are price insensitive and expects high prices to add 2 Mt of supply next year.
On the LME, Hartree Partners built the largest position in October copper, long October against short November, per Bloomberg. The trader is Sonny McNess, Hartree's head metals derivatives trader since leaving Mercuria this year.
The Oct-Nov spread hit $91.99/t backwardation before Tuesday's expiry. The trade pays if US refined copper tariffs squeeze nearby supply, since much LME stock sits in US warehouses.
NYC Hiring - “What is in it for me today”
SV Hiring - “How much the stock options will be worth after vesting”
Ash | Carry retweeted
Thinking about entering tokenized markets? Make these 5 decisions first:
Why: start from a real business problem
Who owns what: legal rights, KYC, AML, sanctions
How it runs: every step needs a named owner, keys included
What proves it's ready: test like it's TradFi
How the tech fits: networks that plug into what you already run
Tokenization changes how assets move, not what they need.
A 24*7 worker for $200/Month without complaining or overtime is a modern day wonder.
Thanks VC dollars for making this possible.
Ash | Carry retweeted
AI Compute is powering the next digital revolution.
From GPUs and data centers to cloud computing and AI inference, compute is becoming the backbone of the AI economy.
Explore the future of AI compute markets with Carry.Exchange.
Ash | Carry retweeted
Two things can be true:
(1) demand for compute continues to increase
(2) lower tier neoclouds fail
This can be reconciled by realizing that the vast majority of compute demand is serviced by the hyperscalers and tier one neoclouds. The long tail of neoclouds mainly serves startups and other thinly capitalized orgs. If those customers fail—and startups rise and fall in unison—you can easily have a situation where compute demand continues to increase and many neoclouds fail.
This 👇 tells the story that most of compute derivatives startups get wrong.
GPU rentals derivatives would at best be notable mention. The volume will be in inference markets and not GPU compute.
Who actually buys GPU compute? Very few companies, as it turns out.
9% of @tryramp customers paid a GPU vendor in the past year, against 78% that paid for an AI application. That small group spent nearly 80% as much as everyone buying the apps.
How they buy is more telling. 8% bought compute through model serving and inference providers, 3% through neoclouds, <1% through wholesale capacity.
Buyers would rather pay for finished work than rent a GPU and sort out the rest.
That’s what we at @notablecap call workload clouds and it’s the bet @burkaygur and @gorkem made for generative media with @fal, what @rauchg made with @vercel, and what @pk_iv made with @browserbase.
Compute quants are shocked to discover that A100 are fully depreciated on the books but are still humming smoothly.
This is what happens when compute derivatives startups get funded without any founder-market-fit.
Go to any emerging market and you will find how textile machines long written off in Europe and USA are imported and run for another 25 years producing clothes exported to US/EU stores.