🌐RTN: why OpenAI needs to build hardware to become a platform
+ automated shipyards, Vinod Khosla on climate investing, the "green premium" myth busted, emissions in the defence industry
OpenAI to build hardware with Apple’s Jonny Ive?
You may have seen the story about OpenAI’s Sam Altman, approaching famed iPhone designer Sir Jony Ive about a potential OpenAI mobile handset. Some believe that LLMs have reopened the market for consumer hardware such as mobile, though I’m less convinced. Critically, hardware is a maligned and underappreciated moat that all of the largest tech players need - is OpenAI next?
Hardware as a moat
The largest software platforms, need to vertically integrate by building hardware in order to create stronger moats:
Google → search → Android, GCP, Nest, Waymo
Amazon → e-commerce → AWS, Alexa, Amazon Logistics
Facebook → social → Oculus
Microsoft → software → Xbox, Hololens, Azure
The reason for this is, controlling the hardware allows you to control the software. Nowhere is this clearer than with Apple’s iPhone. Apple controls (nearly) all downstream functions of the mobile user experience; the OS, your apps, payments, biometrics, and even who can repair your phone.
In fact, a look at the top 10 global technology companies, suggests approximately 60% of their revenues come from hardware - cloud, consumer electronics or chips.
The notable name in this list is Meta, who have diminimous hardware reveneu and have been heavily investing in Oculus, and recently announced smart glasses. Meta, is a good example of the dangers of not having hardware and an OS - 60% of their US users use Apple’s iOS. Meta is downstream of iOS and in 2021, Apple decided to change their data privacy policy. Meta could no longer access the user data it needs from handsets to serve ads - this was a $10bn revenue impact.
Back to OpenAI - do they need to build hardware now? They can build both custom GPUs to train models and a mobile phone for consumers to use those models.
Competition amongst LLM companies is fierce, there’s a land-grab for GPUs and differentiating amongst chat models is quite subjective. We’ve seen other platforms starting to build their own silicon such as Tesla’s Dojo and Google’s TPUs. There have been a few attempts at AI-native consumer electronic form factors, like Humane’s AI pin, Rewind’s Pendant and Avi Schiffman‘s Tab.
Colour me sceptical, but I suspect this may be fundraising voodoo from OpenAI. Taking on Apple and Google (Android) who have 98% market share, their own AI capabilities and ingrained user behaviours will be close to impossible (though I wouldn’t bet against Sam Altman).
Currently, OpenAI is downstream of all hardware and operating systems. If it wants to truly become a platform it will need to build hardware, whether that’s a mobile phone handset, custom silicon or a more novel consumer form factor, remains to be seen.
⚡️Energy and Climate
Interview with Vinod Khosla on investing in climate technology (Azeem Azhar)
Khosla Ventures is one of the most active investors in climate and frontier tech - even in Europe
Vinod Khosla is one of the most successful entrepreneurs and VCs, having founded Sun Microsystems, Juniper Networks and was a General Partner at Kleiner Perkins prior to founding Khosla Ventures
H2 Green Steel is already seeing customers paying a premium for ‘green’ steel (Argus Media)
The “green premium” is an oft-vaunted and often disproved economic concept
The Swedish company who are building a green steel manufacturing facility expects to produce steel with an order of magnitude fewer CO2 emissions, by using renewable energy and green hydrogen
The company’s CTO, claims that they are seeing a 20-30% premium compared to brown (blast furnace-based) steel prices from customers. It says it has already lined up customers including BMW, Mercedes, and Cargill
Steelmaking contributes to 7-9% of global greenhouse gas emissions
Growing Forests: as countries develop, deforestation drops (Works in Progress)
Top Stories
NuScale Nuclear signs contract for SMRs for data centres in US (NuScale)
Open-X Embodiment is the largest open-source real robot dataset to date (various academic institutions)
Allied protection, requires Allied software (Breaking Defense)
Research suggests EVs might actually require more workers than for ICEs (Heatmap)
Tesla’s Texas lithium refinery is coming online sooner than expected (Electrek)
Europe’s first Exascale computer has been signed off (EuroHPC)
Russia to use banned H100 chips to built Top 10 Supercomputer (Tom’s Hardware)
3D Printer uses levitating magnets to break speed records (Tom’s Hardware)
Staff from ARM’s China subsidiary are quitting to create a new govt backed rival (Bloomberg)
Saronic, an autonomous ships startup, raised a $55m Series A (TechCrunch)
🌍 Policy and Geopolitics
Defence Zero: a report on emissions from the defence sector (Roland Berger)
The defence sector emissions are roughly 1-2% of total GHGs, and comparable to that of civil aviation and maritime transport
These emissions are roughly distributed as; Air Force (~20–25%), Navy (~30–35%), Army (~30–35%)
“the Defence sector releases to the atmosphere emissions on the order of 500 megatons to 1 gigaton of CO2 equivalent annually, implying that it contributes between 1% and 2% of global anthropogenic emissions”
Supply chain: an interesting second-order effect of the transition to Net Zero, is how OEMs and the supply chain are affecting the construction of military hardware
An example is, the transition from mainly metallic fighter jets to the use of composites, which reduced the operational emissions of that hardware by roughly 50%
🚖 Moving things
Great thread on some of the innovations and vertically integrated hardware that Zipline want to bring to market
Zipline pioneered the use of drone for deliveries in various countries in Africa where infrastructure is less developed, they have done 40 million miles of autonomous flight (at Sept ‘23)
AI, Digital Twins and Autonomy for use in modernising shipyards (Navy SBIR)
The US Navy is seeking to modernize and improve the efficiency of its 4 public shipyards through technologies like AI, autonomy, and digital twins
Key focus areas include:
Materials handling automation - Automating cranes, forklifts, and logistics tracking with AI/ML to enable 24/7 optimized operations.
Autonomous 3D scanning - High precision robotic scanning and AI-enabled processing to build detailed digital twin models.
Autonomous inspection - Automating non-destructive inspection techniques like ultrasonics and magnetics with robotics and AI
The US military is looking to bring huge amounts of automation and intelligence across all its infrastructure
This is a large opportunity for private companies in; tele-ops (Ottopia, Phantom Auto), warehouse intelligence (Dexory, Terminal, DroneScan), automated warehouse robots (Locus, Fetch) and autonomous heavy industry