Codes of Culture | Issue 91
From chips to culture, the infrastructure race is on.
Welcome back to Codes of Culture. I’m Ashumi Sanghvi.
I’ve been thinking about the word declared this week. Jensen Huang didn’t announce a product at GTC. He declared something. An age. A new unit of industrial organisation. The kind of statement that tends to sound grandiose in the moment and obvious in retrospect. I’ve been paying attention to that register lately, because I think we’re in a period where the declarations are arriving faster than the frameworks to receive them.
Amy Webb staged a funeral at SXSW. L Catterton moved $313 million into Japan. A startup backed by Antoine Arnault is solving fashion’s oldest problem with physics.
None of these is a trend story. Each one is a signal that the systems underneath the systems are being renegotiated. That’s the thread this week.
Remember to subscribe for full access to our news, insights reports, and global events. We are currently working on being at F1 in Miami, the Milken Global Conference in LA, and Cannes Lions. Reply to this email if you will be there too and want to stay in the loop for what we get up to.
📖In this issue:
Jensen Huang declared the age of the AI factory. NVIDIA is no longer a chip company.
Humanoid robots are on the production line. The physical AI inflexion point has arrived.
Amy Webb staged a funeral for the annual trend report. The intelligence format itself is being redesigned.
L Catterton is deploying $313 million into Japan. The luxury capital map is being redrawn.
NVIDIA and Antoine Arnault are backing virtual try-on. Fashion’s fit problem finally has serious investors.
Jensen Huang declared the age of the AI factory. NVIDIA is no longer a chip company.
What’s happening: At GTC 2026 in San Jose on 16 March, NVIDIA CEO Jensen Huang delivered a two-hour keynote to more than 30,000 developers at the SAP Centre, reframing the company’s identity in full. The strategic claim is now full-stack AI infrastructure: inference, agentic systems, AI factories, physical AI, and robotics. Huang announced that cumulative orders for the Blackwell and Vera Rubin chip architectures are projected to reach $1 trillion through 2027, doubling the $500 billion estimate issued six months prior. The keynote’s five themes were inference, the AI factory, the NemoClaw agentic framework, physical AI, and robotics. Huang closed by observing that he could not think of a single company building robots that was not working with NVIDIA.
TLDR:
$1 trillion in cumulative orders through 2027, doubled from prior estimates. Demand for AI compute has not plateaued; it has steepened.
The AI factory is the new unit of industrial investment: purpose-built compute environments for continuous AI production. NVIDIA is positioning itself as its own operating system.
NemoClaw: NVIDIA’s enterprise-secure agentic AI framework for long-running autonomous agents. The shift from prompts to persistent agents is now an infrastructure-level event.
Physical AI and robotics: Huang noted that every major robotics company is building on NVIDIA. The Omniverse simulation-to-deployment pipeline is becoming the industry standard.
The Groq 3 LPU, from NVIDIA’s $20B Groq acquisition, ships Q3 2026 as a dedicated inference accelerator alongside Vera Rubin. NVIDIA is closing the inference cost gap against in-house hyperscaler chips.
Why it matters: The GTC keynote is not a product announcement. It is a positioning statement about where industrial value is accumulating in this decade. For the founders, operators, and capital allocators in our network, the signal is that NVIDIA’s moat is now architectural: it spans compute, simulation, and agentic deployment. The organisations building on that stack are placing a structural dependency they may not fully see yet. Understanding what it means to operate inside the NVIDIA infrastructure layer and where the leverage points sit is becoming a strategic question, not just a technical one.
Humanoid robots are on the production line. The physical AI inflexion point has arrived.
What’s happening: Schaeffler, the German precision engineering group, has deployed a humanoid robot called Digit at its South Carolina plant to handle repetitive material-handling work previously performed by humans, one of the first documented deployments of a humanoid in a live, mainstream industrial facility. In parallel, Mind Robotics, the industrial robotics company spun out of Rivian by founder and CEO RJ Scaringe in November 2025, announced a $500 million Series A on 11 March, co-led by Accel and Andreessen Horowitz, valuing the company at $2 billion and bringing total funding to $615 million. The company is building a full-stack platform of foundation models, purpose-built robots, and deployment infrastructure, using Rivian’s production data as a training flywheel, with large-scale deployment inside Rivian’s own factories targeted by the end of 2026.
TLDR:
Schaeffler’s Digit deployment is not a controlled lab test. It is a live production environment. The proof-of-concept phase for humanoid robots in manufacturing is over.
Mind Robotics’ $500M Series A is among the largest in industrial robotics history. The $2B valuation comes just months after a $115M seed round. This is structural conviction, not speculative capital.
The thesis is that existing industrial robots handle repetitive, pre-programmed tasks. Mind Robotics is building for dexterous, adaptive, reasoning-intensive work; the majority of factory value-add that automation has never reached.
Rivian’s production data, used as a training flywheel, is a competitive moat that very few startups can replicate. Real factory data at scale is the scarcest resource in physical AI.
NVIDIA’s GTC keynote and Mind Robotics land in the same week. The compute infrastructure and the application layer are scaling in parallel.
Why it matters: Future+ tracks physical AI as one of the forces reshaping the operational economics of industries that touch manufacturing, logistics, and high-touch physical services. Together, the Schaeffler deployment and the Mind Robotics round mark the transition from pilot to production. For the operators and investors in our network-building or backing businesses with physical-process dependencies, the planning horizon for AI-driven operational change has just compressed. The labour-cost calculus and the capital-investment logic for physical automation are being rewritten now, not in five years.
What’s happening: At SXSW 2026, Amy Webb, CEO of the Future Today Strategy Group, opened her session by staging a mock funeral for the trends report format her firm helped define. Guests received tissues at the door; the ballroom was set with sombre music and a slideshow. The argument was blunt: an annual PDF captures a snapshot of a moment in a landscape now shifting too fast to summarise once a year. The new framework replaces discrete trend lists with what Webb calls convergences: tracking what happens when AI, robotics, biotech, energy infrastructure, and geopolitical competition collide simultaneously. Her framing borrowed from meteorology: if trends are individual weather data points, convergences are the storm systems that form when those forces combine. FTSG’s clients include Mastercard, Ford, and NASA; the session played to roughly 1,500 people.
TLDR:
Webb is not updating the format. She is retiring it. The person most responsible for institutionalising the annual tech trend report has concluded it is structurally inadequate for the current moment.
The convergence argument: trends observed in isolation mislead. The value is in tracking where AI, biology, robotics, energy, and geopolitics intersect and compound. A yearly snapshot cannot track a live system.
Webb introduced the concept of the agentic economy: AI systems that plan and act autonomously could shift the internet from search and browsing toward delegation, and the companies running those agents become the new gatekeepers.
Her closing line onstage: the next internet is being built not for people, but for machines. That is a planning problem for every organisation that assumes the consumer is the endpoint.
The position Webb has just vacated at the institution that produces the authoritative annual report is now open. The demand is shifting toward continuous, contextualised intelligence.
Why it matters: The organisations in our network that still calibrate their strategic environment using annual benchmarks are working with an instrument that Webb has now publicly declared obsolete. The competitive advantage is shifting to operators who maintain continuous analytical fluency: reading forces as they move, not as they appeared twelve months ago. Webb’s provocation is also an opening: the space she is leaving behind is precisely the one Codes of Culture and Future+ are built to occupy.
L Catterton is deploying $313 million into Japan. The luxury capital map is being redrawn.
What’s happening: L Catterton, the private equity firm backed by LVMH, is targeting approximately 50 billion yen ($313 million) for investment in five Japanese businesses over the next three years, focusing on cosmetics, food, pet care, and restaurants. The commitment comes as Japan’s private equity market recorded an 81% jump in total deal value last year, even as deal activity across Asia declined 14%. L Catterton has operated in Japan since 2017, investing in nine companies, including denim brand Kapital and the world’s largest Kobe beef restaurant chain. One recent investment is a stake in premium restaurant operator HUGE, which is targeting a doubling of annual sales and expansion across Hong Kong and Singapore ahead of a 2030 IPO.
TLDR:
Five businesses, three years, $313M committed. This is a structural deployment, not opportunistic deal flow: L Catterton is building a Japan portfolio with the full LVMH network behind it.
Japan’s PE deal value rose 81% last year while Asia-wide activity fell 14%. The divergence is not coincidental: Japan is attracting capital precisely because the rest of the region is contracting.
L Catterton’s focus: founder-led and family-run businesses facing succession challenges, and younger companies seeking scale support. Traditional financial investors structurally underserve these.
The HUGE investment signals the thesis clearly: back premium consumer brands with genuine domestic depth, then use the LVMH network to take them into Southeast Asia and beyond.
The geography of luxury capital is now India, the Gulf, Japan, and Southeast Asia. Europe and the US are structurally losing ground.
Why it matters: Future+ tracks the rebalancing of luxury capital as a core intelligence line, and the L Catterton Japan deployment is one of the clearest expressions of a pattern we have been watching: premium capital is following culturally embedded demand rather than aspirational Western proxies. Japan’s combination of consumer depth, a favourable currency environment, and an active succession pipeline in family-run businesses is producing a specific window. For the luxury founders and brand operators in our network with Japan presence or ambitions, the capital is moving toward markets you may already be in.
What’s happening: At GTC 2026 on 16 March, Nvidia announced a partnership with AI startup Catches to launch RealFit, a virtual try-on platform for fashion e-commerce. Catches, Nvidia’s official independent software vendor for the fashion industry, has secured $10 million in investment from investors including LVMH’s Antoine Arnault, model and entrepreneur Natalia Vodianova, former Tommy Hilfiger CEO Gary Sheinbaum, and former BCG luxury head Sarah Willersdorf. RealFit’s differentiating claim is physics: rather than merging two images, the system models how a garment actually drapes, moves, and fits on a specific body, accounting for size, shape, cut, and texture. The technology is already live on Amiri’s website, with further luxury and mass-market brands due to be announced.
TLDR:
Physics simulation, not image overlay. RealFit models how a garment behaves on a body: drape, texture, movement, and fit imperfections. That is a material shift in fidelity, not an iteration on existing technology.
Antoine Arnault is an investor, not a brand partner. Alongside Willersdorf (BCG luxury) and Sheinbaum (Tommy Hilfiger), the cap table signals strategic conviction about where luxury e-commerce is heading.
Already live on Amiri. The technology is in production, not in pilot: more luxury and mass-market brands to follow.
The return rate problem is the highest operational cost in luxury e-commerce. A credible physics-based fit solution restructures unit economics across the industry, and, as Catches notes, could grow the total addressable market, not just redistribute it.
NVIDIA’s involvement means the compute infrastructure for real-time physics simulation is now commercially accessible. The technology constraint is no longer the binding one.
Why it matters: For luxury houses, multi-brand platforms, and e-commerce operators in our network, RealFit is the first virtual try-on product to combine technical credibility with luxury-insider backing. The question is no longer whether physics-based fit simulation works. It is how quickly it becomes the expected baseline and who controls the body data it generates at scale. As Vodianova framed it: Catches does not guess how a garment drapes: it knows. That distinction, applied across millions of fit interactions, will be one of the more significant proprietary data assets in luxury retail over the next decade.
 | A guest post by
|