The column: China and tech

"Who was that transistor salesman?"

Charles de Gaulle, President of France, after meeting Hayato Ikeda, Prime Minister of Japan, 1962.

In the 1950s and 1960s, the general perception of Japanese exports was that they were cheap, low-quality copies. In the 1970s people started to get nervous, and in the 1980s people panicked. China is going through that now. 10-15 years ago most people in tech would have said that China was a good outsourcing centre but that its own products were mostly low-quality copies. No-one says that today: China combines efficient low cost manufacturing with ferocious iteration, innovation, duplication and creation of companies and products - and indifference to IP rights. It's moving up the value chain.

This can be overstated. iPhones are made in China but 80-90% of the value of the components comes from elsewhere; China desperately wants a chip business but doesn't have one; and it often doesn't create the core IP of what it makes. One can argue about the details of this: where do you put China in the value chain for cars or solar, or AI research, and how long will a local chip industry take? But Huawei has become the world's leading telecoms equipment manufacturer and Chinese consumer internet companies are a key place many in Silicon Valley look to try to spot the future. There are, after all, more smartphone users in China than in the USA, Western Europe and Japan combined.

That doesn't mean this is going to carry on looking like Japan in the 80s and 90s, though. 

• First, Japan has half the population of the USA, but China has four times the population of the USA, and on current forecasts will still be more than triple even in 2050. Japan's economic growth meant it joined the club, but if China's GDP per capita closes with the USA in the same way (today it's only 15% on some measures), then, deterministically, China might become a bigger economy than the USA, just as the USA became a much bigger economy than the UK or Germany in the late 19th century. 
• Second, Japan was a pacifist country tucked under the US strategic wing, run (more or less) by liberal, democratic, constitutional politicians who did not want to become a Great Power (been there, done that), and China, obviously, is not. 
• Third, Japanese companies bought movie studios and record labels, but no-one worried that let them influence elections. 

In the last 5 years the crisis of the US political system has coincided with Chinese leadership that has moved the country firmly away from any path of convergence with 'western' models of behaviour (even if that was wishful thinking), and strategic... caution towards the Chinese state has gone from a fringe of hawks that was often dismissed as xenophobia to a mainstream government view around the world. I'm not a geopolitical analyst, and I don't know if we should be expecting the Chinese navy to conduct freedom of navigation operations in the North Atlantic or indeed the North Sea in 10 or 20 years, or if we should compare China not just with the Japan of the 1980s but with the Japan of the 1920s. But we do need to think about how this changes the landscape for tech. 

Very obviously, we have a global integrated supply chain and a lot of manufacturing and final assembly happens in Shenzhen: that now means a lot of companies are thinking about vulnerability both to the Chinese state and to sanctions against the Chinese state. Equally, China must be giving even more attention to its dependence on strategic technology from abroad that might suddenly be restricted. 

More interesting: Tiktok is the symptom of a broader change, or a catalyst for realising it. For most of the last 25 years the internet was American by default - the ideas, culture, regulation and companies came from the US and mostly the west coast. In 2008 the US was 80% of global VC investing; now it's 50%. Silicon Valley is still the global cluster, but it's no longer the only place you can make great products - software is diffusing. And in parallel, the internet has become a lot more important - it's gone from interesting and exciting but not really part of most people's lives to systemically important to society, public discourse and democracy. So what happens if now lots of people love a Chinese product? (A contrast: the Russians tried to subvert US social networks but the Chinese built their own.) Tiktok claims 800m MAUs. The US talks about banning it, but you can't ban every cool new app, and yet we need to be conscious that any Chinese company (or company with people in China) can be told to do anything by the Chinese state, and they don't have a choice.  

So, are we worried about the app hacking your phone and stealing email? That's an argument for Apple's sandboxed security model to be tighter. Or about how your behaviour and interests inside the app are tracked? Then maybe we need more practical and coherent rules on data and privacy - a GDPR that's sensible and that can be enforced repeatably? Or do we worry about harmful content? We're deep in a debate (and sometimes panic) about content moderation of American social products, but for Tiktok the problem might be what it hides, so how do we deal with that? It may be useful to compare this with current thinking on enterprise network security: instead of building a firewall and trusting everything inside, treat every service as untrusted and constrained. That also links into a broader theme of mine - that tech is becoming a regulated industry (which is the topic of my next essay). 

News

Huawei booted from the UK. The UK had previously tried a middle path of allowing some Huawei kit in some parts of 5G mobile networks, ignoring heavy US pressure, but has now said it must all be removed from all mobile networks by 2027. Fixed-line networks are being 'advised' to transition away as well. So: 

• The telecoms equipment market has been massively consolidated in the last 20 years, and Huawei has become one of the biggest players, with on many axes the best and cheapest equipment, helped by massive state support and to a backdrop of persistent stories of systematic IP theft from competitors.
• If the UK doesn't use Huawei, deploying 5G will take longer and cost more. But 5G really isn't a big deal, as I wrote here. 
• There's no transparency at all on who really owns Huawei, but that's a red herring: all Chinese companies are subject to the arbitrary demands of the state.
• There have been regular stories of massive security holes in Huawei equipment: not so much deliberate back doors as huge unpatched bugs - either way there are spying concerns...
• But I'd suggest the real issue is China's newly demonstrated willingness to pick fights. Regardless of whether the equipment itself is secure, do you want a piece of critical national infrastructure to be dependent on the good will of the Chinese state for the next decade or two? The answer to that question today feels very different to what it would have been 5 years ago, and lots of other countries will look at the UK decision and feel the same.
• Link

Indian telco deal-making. In the last three months, Google, Facebook and a group of top-tier PE firms (KKR, Silverlake, TPG, General Atlantic, Intel and Qualcomm Capital and others) bought a total of 33% of Reliance Jio in a series of deals, for around $20bn.

• Reliance is a huge family oil / chemicals / textiles / retail conglomerate; in 2015 one of the founder's two sons, Mukesh Ambani, created Jio as a new 4G network and barged into an already very overcrowded mobile market with terrifyingly cheap prices, seizing first place (and bankrupting his brother, who already had a mobile business). It now has close to 400m users, consuming an average of 11 gig per month each, and paying an average of $1.04 month. Something over 100m of these are on subsidised 'Jiophones' - proprietary data-enabled feature-phones.
• Jio has the usual telco 'value-added services' (messaging apps, Zoom clone etc), but has ambitions to be much more, linking up India's mostly unconnected small traders with digital commerce, and moving its base onto smartphones, which is where Google and Facebook come in. Lots of 'digital India' talking points here. 
• Every five years or so, a big telco thinks it can move up the stack and compete with the internet. This is a little like a municipal water company trying to get into the soft drinks business. Jio may be different: it has a more captive, less sophisticated base, a retail arm to leverage, and maybe more ability to innovate and understand the market. Or maybe not. 
• Regardless, it's a leveraged play on the digitisation of India and a large and very well-connected player in a highly political market. And especially with the changing geopolitical situation, India is probably moving up people's lists. 
• Link

Google's investment in Jio ($4.5bn) was part of a $10bn investment fund that Google also announced this week for digitisation in India. Link

GTP-3. Much excitement over the weekend as people started playing with a new machine learning tool released in beta by OpenAI. It's a generative text model, trained on an unprecedentedly large data set, which means that can give it some text and ask it to generate more. In practise, as the examples of the launch page show, that means you can give it the wikipedia page for bread, ask it 'why is bread fluffy?' and it will generate a paragraph of plausible-sounding text. Or, you can ask it to write code for a simple task and it will do it, sometimes, if you prime it properly. OpenAI's own page is actually pretty vague, but twitter is full of examples  - when it works, it really works. A few years ago I described machine learning as giving you infinite interns: here, 'have a look at this and write something similar'. Link

Softbank selling ARM? The WSJ reports that Softbank may be selling ARM - Masa bought ARM in 2016 for £23bn, a 43% premium to the market price, claiming this fit into a vision of digitising everything on earth with smart chips and IoT. ARM owns the IP for the basic architecture of the CPU in every modern smartphone (and the new 'Apple Silicon' chips in future Macs, and lots of other things), and has a business selling blueprints for other people to make some or all of the rest of the chip as well. That means it gets some money from every device sold, but not actually that much, and there's also not much broader strategic leverage further up the stack. Now Softbank probably needs the liquidity, but in the current geopolitical environment buyers will obviously get careful scrutiny from the UK government, as well as the US and pretty much everyone else. Link ($)

Apple's Irish tax win. Apple won an appeal against an EU Commission claim that the Irish government had charged it too little tax. This means that Apple doesn't now have a bill for €13bn in back taxes. Link

EU v. smart speakers. Meanwhile, the EU is also opening a competition investigation into smart speakers (i.e. is Amazon being unfair to competing services). Makes definition is everything, so if you define the market as 'Alexa' then you will probably discover that Amazon has market dominance, but really, so what? It's hard to see this at the top of anyone's list of problems. EU Commissioners are famously keen on easy headlines, though. Link

Twitter got properly hacked. Full details aren't available yet, but it looks like someone got access to Twitter's own internal tools, and then used that to post messages on some famous people's accounts, saying 'send me Bitcoin and I'll send more back'. Apparently they only made $200k or so, which is a pretty low yield use for that kind of access - imagine reading all the DMs. A bit like the dog that caught the bank truck - once they got in they had no idea what to do with it? Meanwhile, we don't yet have the full story on how they got in, but the NY Times report claims Twitter dropped several balls. Links: Twitter report, NY Times ($)

No Google Cloud in China. According to Bloomberg, Google scrapped a project to extend its cloud computing services to China and other 'sensitive' markets with ring-fencing of some kind. Given how hard it would have been to make this work (and to compete in China without local political pull) I rather wonder if this was just a feasibility study? Link ($)

Substack traction. Andrew Sullivan, a popular New York Magazine columnist, left for Substack. Very US media bubble, but Substack is disrupting Medium so fast it hurts. Link

Remember Google Balloons? In the happy free-spending times of a few years ago Google had a project to beam down internet from free-floating balloons in the stratosphere, steering them by changing altitude to get into different air currents. Well, they're operational, apparently, over parts of Kenya, and got there themselves after launch in the Caribbean.  Link

Reading

Jay Goldberg on Softbank potentially selling ARM. Link

I went on the 'Out of Hours' podcast to talk about running a newsletter as a side project. Link

Interesting Google study on purchasing journeys through search. Link

Mosaic Ventures on remote hiring. Link

NY Times digs into how department store closures will take malls with them. Something like a quarter of US malls are probably at risk. Link

Fascinating Reuters piece on how Mexican cell tower repair crews deal with drug cartel ‘narco-antennas’. Link

Wunderman study on how lockdown has affected UK consumer attitudes to online shopping. Link (PDF)

There are a lot of Indian people in Silicon Valley, and apparently, some of them have brought caste prejudices with them. Link

Useful writeup of the fiasco of the UK's contact tracing app project. Link

Porsche is testing 3D-printed pistons. Link

Interesting things

'Please scream in your heart'. the CEO of a Japanese theme park shows how to ride a rollercoaster without screaming. Link

From the Gibsonian files: electricity is free to consumers in North Kosovo, for local political and historic reasons. So, lots of people have been mining cryptocurrencies - so much so that local consumption went up 20%, and in 2018 the European power grid association, covering 25 countries, complained about frequency and power input deviations that were making electric clocks across Europe lose time. (I wonder if that would have worked at Nakatomi Plaza?) Link

(This was much too good to check, but I checked the first fact anyway. True. Link)

Stats

The story of UK lockdown through Google Trends. 'Gym' down, 'running' up, and more. Link

Lots of stats on email from Mailchimp annual review. They sent 3.2bn emails last year, and their customers have 4bn unique email addresses on file. Link