In the great race to conquer world markets, stock exchanges and tech companies are pairing up. A year ago, the Chicago Mercantile Exchange struck a deal with Google, and Nasdaq agreed to link up with Amazon Web Services. Now, it is the turn of the London Stock Exchange (LSE), which agreed this week to make Microsoft its dancing partner for the next 10 years: the LSE will spend £2.3 billion on Microsoft services and, in turn, Microsoft is buying a four per cent stake (£1.5 billion) in the exchange.
What all of these deals have in common is that they involve financial firms trying to become data companies and, to that end, seeking a tech partner that actually knows what it’s doing. This is especially urgent for the LSE, which spent billions on the data company Refinitiv in 2019 and has struggled to digest it ever since. But essentially all the exchanges are trying to do the same thing: move all their data onto the cloud, make it all work properly and then build an artificial intelligence infrastructure around it.
A stock exchange used to be simply about discovering the right price for a share in the most efficient way possible. The price was the data point that really mattered. But over the years, the humble price discovery mechanism has expanded and started to use and produce floods of data about market and business behavior, risk, compliance, valuations and so on.
Among the nuggets stockpiled by Refinitiv, for example, is a database of 2.7 million senior company officials, with pay and employment history going back 24 years; a platform tracking every single project China builds in 70 countries through its $500 billion “belt and road” investment scheme; and a record, updated daily, of the buying behavior of thousands of American grain elevators (the modern version of a granary).
But how well does any of this really work and what on earth is one to do with all this data? That’s where the tech companies come in. They are supposed to build tools that make all of this useful and usable in real time from remote locations. Easy, right? What could possibly go wrong? Still, let’s not fault them for trying. We need investment and efficiency improvements. This is the sort of deal that generates them.
The race to crunch data is now a feature of almost every modern industry, including medicine. I recently listened to Prof. Rick Stevens, a computer scientist at the University of Chicago, explain how he is trying to use a supercomputer to identify new treatments for cancer. The computer, at the Argonne National Laboratory, is being fed with all sorts of data about cancerous tumors and outcomes from clinical drug trials. From this, the scientists are trying to write algorithms that can match up particular types of cancer to new drugs or drug-combination treatments that may not have been tried on them. In the past five years, Prof Stevens told me, there have been more than 100 research papers published taking similar approaches.
There are several severe limiting factors on this work, however. The first is computing power. Very few computers are fast enough to make use of the amount of data being fed into Argonne and very few facilities can make the sort of parts, like advanced semiconductors, that they require. Beyond that, there simply aren’t very many people capable of building and using such a tool. And if one could suddenly build thousands of these computers and train up millions of people to use them, we would run into a new limiting factor: the huge amount of energy they use.
This, though, is the future. Both the US and the EU have announced their intention to fund multi-billion dollar projects building new, onshore semiconductor supply chains to reduce reliance on Taiwan. The UK is dilly-dallying. We could, in theory, become a powerhouse of computing and engineering talent, given our strong university sector. But the relevant courses seem to be of interest almost exclusively to Chinese students, most of whom go straight home after their degrees. Meanwhile, British students pay as much to study computer science as they cough up for a course in drama and theater studies. It shouldn’t take a super-duper advanced algorithm to figure out that this doesn’t make sense.
.