Welcome back to the Pinsent Masons podcast, keeping you up to date with the latest business law news developments from all around the world. I'm Matthew Magee, a journalist here at Pinsent Masons, and this week we look at the surprisingly complicated ins and outs of AI infrastructure financing and how some of the investors' interests could conflict with each other. And we hear why we're at a crucial inflection point for floating offshore wind generation. But first, some news.
UK grants consent to drill Jackdaw gas field, body stigma and depression content earmarked for stricter UK regulation and HMRC to intensify large business scrutiny after PAC tax compliance reforms.
Gas from the Jackdaw field in the North Sea is expected to be supplied to properties in the UK this winter after consent for the project was granted by the North Sea Transition Authority to operator Adura last week. An earlier consent granted for the project was previously ruled unlawful. Adura said the decision will enable gas from the site to be supplied within weeks. It estimates that at peak production, Jackdaw could produce enough gas to heat the equivalent of 1.4 million UK homes. Energy expert Chris Sawyer said Jackdaw's approval is a pragmatic decision by the UK Government to meet UK's current energy needs. Domestic production can help strengthen energy security, generate tax revenue, support skilled jobs across the UK supply chain and reduce reliance on imported energy.
Online service providers would have to more closely control the extent to which children can access body stigma or depression content on their platforms, under proposals put forward by the UK's online safety regulator. Ofcom has recommended a recategorisation of body stigma and depression content under the UK's Online Safety Act with a view to ensuring greater protections for children against that kind of content. It would be recategorised to priority content that is harmful to children. Online safety regulation expert Gemma Erskine said the difficulty with categories like body stigma and depression content is that they're not easy to define. A post about eating disorder recovery and a post promoting disordered eating can look incredibly similar even with opposite intentions. Tech companies will need to make those distinctions consistently at scale and under the threat of significant regulatory sanction, which is no small ask.
Large UK businesses should prepare for more intense scrutiny from tax authority HMRC after the UK government accepted recommendations on large business tax compliance made by a parliamentary committee, experts have said. The Public Accounts Committee committed HMRC to review its use of enforcement powers, accelerate compliance activity through greater use of technology and data analytics and assess whether elements of its cooperative compliance approach should be extended beyond UK's largest taxpayers. Tax law expert Ian Robotham said the introduction of digital action plans for all large business enquiries is operationally significant. The explicit reference to using information powers to accelerate enquiries signals a more assertive approach where voluntary cooperation is not forthcoming. The committee had expressed concern that HMRC has never formally placed a large business into special measures since the regime was introduced in 2016. The framework is designed to apply enhanced oversight and governance measures where a business is judged not to be cooperating with the authority.
It's easy to get slightly number dizzy reading about the astronomical sums being raised by AI companies to build the data centres, networks and infrastructure needed to deliver the ambitious services they've promised to users. Investment bank Goldman Sachs thinks that $1 trillion of AI related investment will happen this year.
But where's the money coming from and how does this kind of investment work? It turns out it's a lot more complicated than you might think, and that's because AI infrastructure is made up of different kinds of assets with different kinds of investor, Shenzhen based project finance expert Kanyi Lui told me.
Kanyi Lui: Talk about AI infrastructure. Most people think about a room of servers, right, or a room of computers, but it's so much more complicated than that. For an AI infrastructure data centre to function, you need the building, you have the servers or, you know, the GPUs, which are quite a different set of assets these days. You have the power connection. In some of the newer projects, you have renewable energy projects attached to it. You might have energy storage projects attached to it. We've recently heard about a nuclear adjacent data centre, right? So, I mean, that's really interesting. So those are some of the variations that we could be talking about.
Matthew Magee: And these different asset types have different ways of being funded and different funder expectations.
Kanyi: The assets have very different economic lives. So, the buildings, the grid connection, the cooling system, the energy supply could be useful for decades. The GPUs, on the other hand, I mean they can depreciate and be at the end of their economic useful life within two, three or four years depending on who you talk to. So, the mismatch means, look, you know, this is something certainly the investors and lenders have to think about when they're putting together the financing, but also, they have to keep an eye on. I think one of the really interesting challenges for AI and tech is how quickly things move. So, for example, NVIDIA and AMD recently moved from a two-year release schedule to a one-year release schedule and that's impacted on how the residual value of chips have been calculated. So, I think those are all things you have to take into consideration.
Matthew: So, if you imagined one big monolithic investment into a single thing called a data centre, you'd be wrong. These are different investors used to different norms for different kinds of asset classes, and this means different expectations and therefore the potential for clashes, says Kanyi.
Kanyi: I think for AI infrastructure, the ultimate goal is that the financing must follow the asset. You have long term capital that's financing the building, the power generation, the grid infrastructure. Typically, you have shorter term capital or equipment finance, you know, for the GPUs, the technology, the other things which need to be upgraded, you know, much more quickly. And then you have a need for flexible capital, right? So, this is capital that's required to fund the general daily operations, things that might be required for future tech refreshes. So, this could come in the form of equity, some of the debt or other things or leasing, right. But I think different financiers are looking at different assets and have different clocks running. So, a project lender will be looking, you know, to sell the business if things go wrong as a going concern because that's usually a better outcome than selling off, you know, chopping off limbs and selling them off one by one. The GPU financing, on the other hand, they want to get in very quickly and get the GPU out because every day you sit on these things, they lose value. And then the bondholders, I mean, might be a completely different class of creditors, they might be exposed to the group as a whole rather than the asset or the project itself, but with less understanding as to what they're financing, right. So this is typical of the bond market. I think these differences can be very difficult, you know, even when things go well, but they become much more important when things go wrong.
Matthew: Kanyi says the complexity of what goes into a data centre, the interdependency between the parts and the different kinds of capital that go into funding it gives rise to two kinds of greater risk than faced by other big projects, completion risk and concentration risk. First, completion risk.
Kanyi: What I mean here is that the building might be physically complete, but the connection may be delayed, the GPUs may be unavailable or caught in customs or shipping, the software may not be supported, the customer might not be ready. So, in those cases, then you know the project might be there, but you cannot produce revenue. But in one of the cases we've seen in the US, you had a financing via bonds with a backstop from a major hyperscaler. And what happens is that the backstop only activates if the construction is complete.
And if the construction is not complete on schedule for more than 180 days, then the backstop falls away. So, what might look like a supportive financing could all of a sudden very quickly become unsupported with the bondholders or, you know, banks in the parallel universe taking full project risk. Concentration risk is not unique to infrastructure. I think for many large infrastructure projects, you have a scenario where you only have a few buyers, right? And the market's only so big. But if you take a look back at the current AI landscape, what you'll find is that what looks like a number of separate projects on the outside may actually be an interconnected economic system. So I think the concern is really having what looks like diversification at the project level, but not diversification at the system level. I think the revenues and credit support tend to come from a few very large players, right? And these large players, you'll see many of these operate on similar assumptions and cash flows that actually go round, you know, in an interconnected web. So I think the issue then there is if one of them gets into trouble, how would that be dealt with? And that is one of the risks that I think, you know, lawyers tend not to worry about, but I think for lenders that's certainly what we're taking a look at.
Matthew: When this much capital is at stake and at this stage there are worries that effectively the whole world economy is overexposed to AI investment, the debt markets are careful and maybe they have cause to be, says Kanyi.
Kanyi: So if you look at the LLM token expenditure index, AI token pricing is actually quite quickly falling. You know, not consistently, it's gone up and down. But if you look at where things are now in October, you know, it's actually less than half of its peak during the summer, right? So, in the summer, it's over $2.00, now it's $0.94. So, this doesn't mean I think AI demand's falling. I think nothing can be further from the truth. But I do think what that means is that more efficient and cheaper AI, especially those coming out of China, is helping drive down prices. So, I think this is great news for users, but I think it creates a much harder question for investors and bankers. So, will usage grow more quickly than reductions in the token price? Who knows, right?
So, I think that is one of the difficulties of pricing the cost of financing. Last week, the yield on one of the AI related bonds went from 9.75% to 15.78%.
Matthew: Remember, when bond yields go up, that means the value of the bond has dropped. Back to Kanyi.
Kanyi: So, when you have that, that means, you know, there's a lack of confidence and the five-year credit default swap rose by over 1,000 basis points. Credit default swap is people betting on the company will go under. So, if you're trying to buy insurance against the company going under and that reinsurance premium is going through the roof, somebody's saying, look, this is getting awfully risky. So, I think what it means is that the markets will probably just look at things a lot more closer than before and project assumptions will have to become less optimistic.
There are easier and harder ways to generate electricity from wind, for pretty obvious reasons. It's easier on land than on water, and it's easier with turbines that are attached to the seabed than ones that float on the surface. So, it's no surprise that investors have turned to onshore and fixed bottom wind first. But floating wind can operate in deeper waters and access bountiful wind energy not available any other way. It's now an industry at a crucial inflection point, when some of the first big projects are being watched closely to see if they really are commercially viable. There is a lot at stake as energy policy tries to juggle geopolitics inspired questions of security of supply by generating more green energy with the need to keep prices down for citizens.
Edinburgh based renewable energy expert Niamh Deery can help steer us through these profitably choppy waters, and I started by asking her what kind of potential there is for floating offshore wind in the UK.
Niamh Deery: At the moment I think the UK is really one of the most exciting places to be for floating offshore wind. I mean, if you're looking at Scotland alone, we've got about 14 gigawatts of floating wind projects progressing through the planning system as we speak. So that gives you hopefully an indication of the scale of the potential pipeline that's there. You know, people obviously working in Scotland and within the UK know that, but I don't think we're the only ones that have noticed that potential because, you know, the international presence in Scotland and in the kind of Scottish project scene, that's also a pretty striking thing. And, you know, we've got companies from Japan, we've got companies from Korea, Norway, Netherlands, France, and they're all kind of really keen to be involved and kind of get a slice of the pie in Scotland because there's so much going on. And, you know, these companies are, some of them are in the supply chain, some of them are investors. And I think that just reinforces that the UK is a really exciting place to be at the moment.
Matthew Magee: That potential can only be realised with significant investment and support from the public purse, a phase every new energy technology goes through. But there have been other challenges.
Niamh: These offshore wind developers have to contend with the same wider pressures that are affecting the rest of the energy industry. You know, we've got inflation, there's supply chain constraints, we've got higher financing costs, and of course, the geopolitical uncertainty that underpins all of it. But I think while many take a pessimistic view of progress over the last few years, I don't think it's right to be pessimistic about it at all. I think there's still loads of really exciting things happening. And I think a lot has happened over the last couple of years. And, you know, lots of progress has been made. I think one of the kind of big, important changes is actually the nature of the conversation. So, if you think about a few years ago, the question that was being asked was whether floating wind technology actually could work, but not nowadays, the question is more about how do we make floating wind projects commercially deliverable at scale. So that's a really important and exciting transition because we're not focusing on whether or not the technology itself works. What we're focusing on now is can you get financing at the right cost? Can you get your contracts in place? Do the kind of ports that are in the area of the development support, can they support that project moving forward and kind of can the regulatory framework set up in the UK support these projects as well? So I think these are all great questions to be asking and shows that the industry has developed.
Matthew: There are three Scottish projects in particular that investors and developers from all over the world are watching, says Niamh, to see how the transition to full commercial operation works.
Niamh: There's also really a sense that people are watching what's happening in the UK to understand how the floating wind industry can actually be built at scale. And kind of ultimately, I think what they're looking to do is take that experience and try and replicate it in their own domestic markets. The projects that everybody's got their eye on at the moment are Green Volt, Pentland and Erebus. So, they're really seen as kind of the vanguard of UK floating wind projects because they're the ones that are moving towards investment and delivery. And they're the closest to investment and delivery. The rest of the industry is to some extent waiting for proof that these projects can be delivered kind of at a commercial scale. So, you've got developers and lenders and insurers and the supply chain. Everybody's sitting on the sidelines kind of looking at these projects and looking to see that they can be procured and financed and built successfully. Because if these projects can prove that, then what you're also proving is that there's kind of sufficiently predictable returns on the effort involved in developing one of these projects and kind of making this big capital investment. Everybody wants kind of the UK projects to succeed because it'll pave the way for a model that can be used in other countries in the future.
Matthew: Floating wind is more expensive now, but that won't always be the case. Industry estimates say that it will be the same price as fixed bottom offshore wind by 2050. But to attract investment at rates that allow for profit making, developers have to pursue the reduction of costs in the same way other technologies have, through standardisation, economies of scale and automation.
Niamh: I don't think the problem is a shortage of capital. I think, you know, the capital is there. I think kind of what's difficult about floating wind is that what you're trying to do is provide investors with kind of enough certainty around risk and revenue and deployment and make sure that you're attracting that capital on terms that will make the projects economically viable. So, you know, cost reduction there. There's not a silver bullet.
It's really about a combination of kind of standardisation and automation and developing technology and kind of industrialisation and all those things are really closely connected. So, kind of, for example, you know, design convergence is going to be increasingly important for floating offshore wind. So obviously innovation in the sector is really valuable, but the more you can standardise over time how these projects come together. So how the fabrication is done, how the installation is done, how the certification process works, how they're maintained, kind of ultimately what you're doing there is creating a more predictable proposition for investors and for projects in the future. They can just follow that approach rather than all of the different projects having kind of novel concepts that make the financing piece a bit more uncertain. What comes first, the pipeline or the infrastructure that supports the pipeline? They kind of need to come together at the same time so that the risk is minimised for all involved.
Matthew: Like most renewable energy production, floating offshore wind is dependent on an ageing and creaking energy grid to get its power to market. There are no quick or easy solutions to that. But the UK government will publish its Strategic Spatial Energy Plan early next year and the floating offshore wind industry will hang on its every word, says Niamh.
Niamh: What everyone's hoping that's going to do is set out kind of a big picture map for Great Britain and sort of setting out where we should be generating and storing energy in the future and kind of how much is needed in a certain area and kind of broadly when is that need going to crystallise. So having visibility about the kind of overall picture and kind of the overall plan in Strategic Spatial Energy Plan, that's going to give kind of developers real visibility and kind of it has real commercial value to developers because it goes back to kind of that question of certainty in the pipeline and making sure that this is kind of, we've got everything in place to make sure that this industry can be supported in the longer term.
Thank you again for listening, for joining us. We know that there are lots of calls on your time, so we appreciate every minute you spend listening to this. Please do send it on to other people you think it might be relevant to and you can rate or review it where you get your podcasts. You can keep up hour by hour with news and analysis from our experts and delivered by our specialist team of journalists any day of the week at pinsentmasons.com. Or you can get a personalised digest of what matters to you by going to pinsentmasons.com/newsletter and signing up. But for now, thanks again for listening and goodbye.
The Pinsent Masons Podcast was produced and presented by Matthew Magee for international law firm Pinsent Masons.