Perspectives

Grid capex in the age of renewables and AI

Download the PDF

03 July 2026
By Audrey Lee

Executive summary

The grid is the bottleneck. With the ongoing electrification of the economy, electricity networks are becoming a constraint on the energy transition and, in some cases, slowing down the development of the digital economy. An aging electricity grid not designed for modern-day digital demands is being stretched by AI-driven electricity loads and long lead times for grid connection approval, resulting in curtailment of renewable output due to capacity constraints.
An investment super cycle in the offing. Global total grid spending is forecast to reach $US5.9 trillion over 2026–2035 period. In developed markets, aging power assets and compounding electricity demand are creating a multi-decade investment cycle. In emerging markets, millions of kilometres of new lines must be built from scratch.
The opportunity spans the value chain. Transmission faces the most acute capacity shortfall. Distribution networks need both physical and digital upgrades. And the specialist service providers that deliver the work are becoming a critical enabler of the entire programme.

The global electricity grid is entering its most consequential investment cycle in a generation. Across geographies and market structures, the same challenge is emerging: grids need to become more robust, more flexible, and significantly larger. The drivers vary, from renewable integration in Europe to AI-led demand growth in the United States, but they point to the same imperative. The grid, long treated as stable background infrastructure, has become a bottleneck on the energy transition, the digital economy, and industrial growth. Remedying it will demand capital across the entire value chain: in transmission to connect remote generation, for flexibility services to manage intermittency, and on smarter distribution networks for decentralised, bidirectional power flows.

Drivers of the grid investment cycle

The supply side: renewables and the rewiring of supply

Renewable generation capacity is scaling rapidly, but the grid infrastructure needed to deliver that power reliably is not keeping pace. Globally, 814 gigawatts (GW) of new solar and wind capacity was added in 2025, up 17% on the prior year, and enough to surpass four terawatts (TW) of total installed capacity for the first time.¹ What is needed is not just more grid but also a fundamentally different kind of grid. The old model delivered power in one direction, from large, centralised plants to often passive consumers. The emerging model must handle distributed, variable generation feeding in from all directions, from utility-scale wind farms to rooftop solar, while maintaining the stability and reliability that consumers and industry depend on (Figure 1).

Figure 1:
From one-way to many-way: how power flow is changing

Source: Macquarie Asset Management analysis.

The gap between these two models is already visible. In many jurisdictions, grid infrastructure has simply not kept pace with the generation buildout. The result is growing generation curtailment: clean energy being produced but wasted because the grid cannot absorb it or does not have the necessary infrastructure to transport it. In some markets, curtailment rates for wind and solar have risen sharply (Figure 2).

Figure 2:
A record amount of wind and solar power was curtailed in 2025

Source: BloombergNEF (BNEF), “Wind and Solar Curtailment Reach New Records”, April 2026. California Independent System Operator (CAISO) operates California's wholesale power grid.

This problem is often compounded by the geographic mismatches that occur when generation is sited far from demand centres. Offshore wind sits off coastlines distant from industrial areas that need them most, as in Italy and Germany. Solar potential is highest in arid regions, not in the cities or built-up areas that consume the power.² Bridging these gaps requires transmission infrastructure that, in many cases, does not yet exist or is severely under-developed or maintained.³

The challenge goes beyond capacity and is created by the switch to different types of technology. The grid is losing a built-in source of frequency stability as wind and solar rapidly displace conventional power plants across markets from the UK to the US (Figure 3). The heavy spinning turbines these plants use — and the physical inertia this rotational mass provides — acts as a physical buffer, resisting sudden changes in frequency when supply and demand fall out of balance. Every synchronous generator that goes offline takes some of that buffer with it. The April 2025 Iberian blackout, which left tens of millions without power for up to 18 hours, underscored the consequences when grid infrastructure and operational practices fail to keep pace with a rapidly changing generation mix.⁴

Solutions are emerging from grid-forming inverters to synthetic inertia, but they raise the technical and commercial complexity for grid management.⁵

Figure 3:
The displacement of synchronous generation is a global phenomenon

Source: Ember Energy, “Yearly electricity data”, accessed 24 April 2026.

The demand side: AI and the return of load growth

On the demand side, the picture is shifting just as fast. Electrification of transport and heating, industrial reshoring, and the rapid buildout of AI infrastructure are all placing new demands on grids simultaneously. Of these, AI stands out for the speed and concentration of its impact. In the United States, electricity demand is inflecting upward after roughly fifteen years of stagnation. In Asia, where demand has been growing rapidly, data centres are compounding the strain on grids that have struggled to keep pace. From 2025 through 2030, global data centre capacity additions are forecasted to be between 69 GW and 141 GW, reflecting an 85% to 174% growth relative to global capacity in 2024.⁶

The physical footprint of AI infrastructure is scaling rapidly. Average data centre sizes have grown from 25 megawatts (MW) in 2020 to nearly 80 MW for those entering service in 2025,⁷ driven by the computing intensity of large language models and other AI workloads. Connecting these larger facilities often requires new transmission infrastructure well beyond the immediate point of connection, and the queue to secure that connection is growing fast. In Texas’ ERCOT⁸ alone, 144 GW of data centre capacity is awaiting grid interconnection, against a system peak demand of just 86 GW (Figure 4). The scale of demand pressure is clear, even accounting for the high dropout rates typical of such queues.

The resulting capacity constraints in established hubs are pushing development toward regions where land and grid headroom remain more available — in Europe, for instance, toward the Nordics and southern part of the continent.⁹

Figure 4:
Demand connection queues at transmission level in selected markets

Source: BNEF, “Grid Investment Outlook 2025”, December 2025.

Current grid investment plans may not fully reflect the scale and nature of AI-driven demand. As facility sizes continue to expand, the transmission investment required per data centre is materially higher than backward-looking assumptions suggest. The load profile also adds complexity: as Figure 5 illustrates, facility-level power draw varies significantly with utilisation and workload mix. In this example, at 80% average utilisation, a single facility can see intraday swings of 20–40%, creating balancing challenges that compound as data centres account for a larger share of total system load.¹⁰ These variability patterns sit outside what conventional dispatch and reserve planning frameworks were designed to accommodate.¹¹ Hyperscale operators' commitments to match new capacity with clean power procurement further compound the challenge, tying AI load growth directly to renewables buildout and grid integration. Together, these factors point to meaningful upside risk to grid capex forecasts on both the infrastructure and flexibility sides.

Figure 5:
AI training loads are volatile: simulated intraday power swings at a 10 MW colocation facility

Source: Vercellino, R. et al., “Measurement of Generative AI Workload Power Profiles for Whole-Facility Data Center Infrastructure Planning”, National Laboratory of the Rockies, April 2026. Utilisation refers to the share of compute nodes actively running workloads at a given time. The three scenarios (40%, 60%, 80%) reflect the range of plausible operating conditions. No facility sustains 100% node occupancy continuously, as workloads arrive and complete at different times and nodes cycle between jobs. Production GPU clusters have been observed to operate at average node occupancy of 65–82% depending on workload type.

Aging grids, missing grids

These pressures are not arriving sequentially — they are compounding. In developed markets, renewable integration and surging data centre demand are landing on transmission and distribution infrastructure that was largely built in the post-war decades and is approaching the end of its operational life. Asset replacement alone would constitute a significant investment need; layered on top of electrification, renewable integration, and AI load growth, it becomes an investment super-cycle.

In emerging economies, the challenge is different in character but no less significant. These markets are expected to account for nearly 80% of global electricity demand growth through 2030,¹² driven by industrialisation, urbanisation, and rising cooling needs. Yet, much of the grid infrastructure needed to serve that demand does not exist. The IEA estimates that emerging markets outside China need to add 2.3 million kilometres of new transmission lines by 2040.¹³ The buildout is gaining momentum: in Southeast Asia, the World Bank and Asian Development Bank have pledged over $US12 billion to support the ASEAN Power Grid initiative, while Latin American markets such as Brazil and Chile are scaling grid investment to integrate fast-growing renewable capacity.¹⁴ Many of these markets also have the opportunity to deploy modern, flexible grid technology from the outset rather than retrofitting it onto aging networks.

The investment opportunity: where the capital needs to go

The scale of investment required is substantial. Over 2026–2035, global grid spending is expected to reach $US5.9 trillion, split roughly evenly between distribution and transmission.¹⁵

That spending falls into three categories: system reinforcements (44%), which upgrade the wider network to handle increased loads and new generation; asset replacements (35%), to retire infrastructure that has reached the end of its useful life; and new connections (21%), to link new sources of demand or generation to the grid (Figure 6).

Figure 6:
Global grid investment in transmission and distribution forecast, 2026–2035

Source: BNEF, “New Energy Outlook 2025: Grids”, August 2025.

The opportunities this creates span the grid value chain, but three segments stand out.

Transmission and interconnection: This is where system reinforcement needs are most acute. New generation clusters and demand centres require long-distance capacity that does not yet exist. In the US, over 2,000 GW of generation and storage capacity sat in interconnection queues at the end of 2025 — 1.5 times the country's entire installed base (Figure 7).¹⁶ This is much larger than 15 years prior. In Europe, cross-border interconnectors represent a specific sub-opportunity: the EU is still lacking 88 GW of cross-border transmission capacity needed to meet its 15% interconnection target by 2030.¹⁷ Across both markets, new generation is being built far faster than the grid needed to carry it.

Figure 7:
Active capacity in interconnection queues, US power plants, 2010 and 2025

Source: Lawrence Berkeley National Laboratory, “Interconnection Queue Dataset & Summarized Data Files, through 2025”, May 2026.

Distribution networks: They are where the energy meets the end user. Electric vehicles, heat pumps, rooftop solar, and urban data centres all connect at the distribution level. In Europe, an estimated 70% of new renewable generation and storage will be integrated at this layer by 2050, further stretching networks already contending with the geographic mismatch discussed above. Yet distribution grids have historically attracted less attention than headline transmission megaprojects. The challenge is compounded by age: approximately 30% of European distribution grids are already over 40 years old,¹⁸ a share expected to exceed 50% by 2030.¹⁹ Beyond physical reinforcement, distribution grids also require digitalisation — smart meters, automated switching, and real-time monitoring — to manage increasingly complex two-way power flows. This is a large and fragmented opportunity set, but one where the investment case is being driven by the same structural forces reshaping the rest of the grid.
Grid services: Delivering on asset replacements, new connections, and system reinforcement requires a broad range of physical capabilities: construction, inspection, power electronics installation, digital monitoring systems, and emergency response. Demand for skilled service providers is growing in lockstep with the capex cycle, and the firms that can deliver this work at scale are becoming a critical enabler of the broader investment programme. The market is fragmented, with meaningful consolidation potential, and benefits from long-term demand visibility underpinned by the multi-decade cycle described above. For a deeper look at this opportunity set, we explore this theme in Infrastructure adjacencies: Where infrastructure meets private equity (April 2026).

Battery storage merits mention as a second-order beneficiary. While not grid infrastructure per se, battery storage directly benefits from the same drivers: renewables integration, grid flexibility needs, and rising demand volatility. As grids become more complex and variable, the value of fast-responding, location-flexible storage assets grows in parallel.

Conclusion

Grids are no longer the infrastructure that is “seen and not heard”; they are the bottleneck for both energy transition and AI-driven load growth. The issues related to variable renewables, electrification and data centres, aging assets, and emerging-market expansion are compounding at the same time. Together, the priority is no longer just building more power plants. It is upgrading and expanding the grid: building new transmission lines, upgrading the technological capacity of local networks to “sweat the assets” more efficiently, and developing the services and control systems needed to connect and run an increasingly complex system. For investors, that means a multi-year opportunity set across interconnection and reinforcement, digital distribution, and the specialist providers that deliver the work.

Podcast

August 2026

Power struggle: Why the grid can’t keep up

For decades, the grid was the most boring, unseen part of the energy conversation — until now. In this episode of Pathways, Daniel McCormack, Head of Research at Macquarie Asset Management, is joined by Audrey Lee, Research Analyst and author of the paper Grid Capex in the Age of Renewables and AI, to discuss why the electricity grid has become the binding constraint on the energy transition, the AI buildout, and industrial growth. They cover the $5.9 trillion investment gap and how it plays out differently across Europe, India, and China, the twin problems of aging and missing grid infrastructure, and how data centre demand is reshaping regulation, tariffs, and the race for behind-the-meter power.

Daniel McCormack, CFA
Head of Research - Host

Audrey Lee
Research Analyst

Listen on Apple Podcasts

Listen on Spotify

Listen on Amazon Music Podcasts

Authors

Intro

Welcome to Pathways, a Macquarie Asset Management podcast where we provide fresh perspectives and insights for institutional investors and consultants about real assets, private markets, and macroeconomics.

Daniel McCormack: For decades, the grid was the most boring and unseen part of the energy conversation. Like the plumbing in your house, it was only noticed when things went wrong. But that's no longer the case. Audrey's paper, Grid Capex in the Age of Renewables and AI, makes the case that the grid is now the constraint on almost everything. The energy transition, the AI buildout, industrial growth. Hi everybody, my name's Daniel McCormack, and I'm Head of Research for Macquarie Asset Management. and today we're going to talk about Audrey Lee's latest paper in some detail. but we're going to touch on three things. One, how different the grid looks by region. 2, what aging and missing grids actually mean on the ground and what they mean for investors. And 3, what the latest is with data centre demand. So, Audrey, great to have you back on the Pathways podcast.

Audrey Lee: Thank you, Daniel. Pleasure to join.

Daniel McCormack: So, Audrey, before we dig into the detail, if there was sort of one sentence or one idea that you wanted the listeners to take away from today's podcast, what is that?

Audrey Lee: Yeah, I would say it's this one idea that the grid is the bottleneck now. We can certainly build the power. Renewables are being added faster than ever to the grid, but what we can't do at the same pace is to move it around and connect it. So now the grid is really the thing that everything else is waiting on. The AI buildout, the energy transition, even industrial growth, everything is waiting for the grid.

Daniel McCormack: I think that's a very good setup because today we're really going to go through the mismatches that are playing out in different places. So, let's just start with a bit of a geographical overview, if that's OK. You know, your headline number from the paper is $5.9 trillion. That is, we need $5.9 trillion of investment in the grid out to 2035. That's a huge number. But it hides a lot of regional variation, and, you know, not just in size, but in character. So, let's start with Europe, if that's OK, because it seems to be the clearest case of the money being there in principle, but politics and financing kind of deciding who actually pays. What's the situation in Europe with respect to the grid and the capex need?

Audrey Lee: Yeah, right, Daniel, I think Europe is a good place to start this conversation. So, I mean, the EU put a concrete number on that in December 25 with their European grids package and the estimate for what's needed to run it is about like, 1.2 trillion for the electricity grid through to 2040. Now, this is really an example where who pays gets explicit. Analysts looking at the financing mechanisms have basically concluded that Europe's grid transition will be paid for by consumers quietly, over many years, because the EU's own funding instruments cover only a small fraction of the need. Now, there's also a live fight over the congestion revenue, like money connected, collected when electricity prices differ across borders. The European Commission wants to earmark a share of it for grid investment, but countries like Sweden, that has already invested heavily in interconnection, see that as being penalised for having done the right thing already.

Daniel McCormack: And you know, just to make it very real for the listener, can you just give us a bit of a mental image for what sort of, you know, an underinvested European grid actually looks like on the ground? So, in the real world, what is that?

Audrey Lee: Yeah, I'll give you some headline numbers. Like, we're talking about 500, over 500 gigawatts of wind power sits in European connection queue right now. And we have more than half of the transmission projects needed by 2030 that are still waiting on permits. And, you know, that's not abstract for ordinary people either. Reports out of the Netherlands described around 12,000 companies waiting, waiting in a grid connection queue, and some new housing developments simply can't get connected because there's no local grid headroom.

Daniel McCormack: Yeah, that's a very concerning thing. A house can be built, but it just can't be plugged into the grid. You know, that's a very real-world consequence of what we're talking about here. Just shifting maybe to emerging markets, and, you know, this is where there's a lot of growth in power demand, you know. Electrification is clearly ongoing, you know. Europe's bottleneck, you know, is money and politics, but if we take a place like India, for example, it's more about timing. Is that fair to say?

Audrey Lee: Yeah, I think that's fair. India's problem is, to a large extent, timing. It is putting up renewables faster than it can build wires to reach them. So, at a federal level, the plan is basically a race to nearly double the grid by 2042. Though the interesting part is who's building it. India is quietly privatising transmission. We're talking about roughly 40% of the new lines that are being awarded are being awarded to private firms. Adani's is the biggest, and you have other like, private firms involved as well. And why that's happening, that's because the companies building solar and wind got tired of waiting for someone else to build a grid. So, they have started building the wires themselves, so that they can control their own projects and get it connected. We have people trying to jump the queue now. You have developers grabbing connection slot years earlier than they really need it, sometimes before the substation even exists. And there's so much of it happening that the regulator calls it 'squatting'. You have about 42 gigawatts of projects sitting on the grid slots with no power contract behind them whatsoever at all. that's why the grid regulator has now started revoking assets to clean the queue, just like some other European regulators have to do.

Daniel McCormack: So, is it fair to say that the grid is so scarce that a connection slot has become the asset?

Audrey Lee: Well, exactly, though I'll be fair to the story, privatising isn't private dominated. State-owned power grid still wins a lot of the work because they have a lower cost of capital. And execution is slipping as well. Of the projects due only about half came in on time, but that slippage is the thesis: the wires just can't keep pace with the generation.

Daniel McCormack: That's fascinating, Audrey. Jumping now to the other major emerging market, China, which is obviously a huge part of the overall equation here. What’s happening there?

Audrey Lee: Yeah, I'm glad that you mentioned China, Daniel. Certainly, like, China is still the largest grid build on the planet. We're talking about 700 billion across its two grid operators over 5 years. But it's just a different kind of story in China. It isn't a market reacting to demand. In fact, quite the opposite. It is a deliberate state program to build, so that it can move renewables and hydropower from the western interior region to the industrial east, through ultra-high voltage lines. And I would say the overall picture is quite similar. It's just muted. In China, you still have someone waits for the grid and you need someone to pay for it. But in the case of China, it's the state that settles both sides administratively instead of through market mechanisms and politics.

Daniel McCormack: Got it. Just bringing this part of the conversation to a close, like that one headline number, 5.9 trillion, which again, is just an absolutely huge number. But if we think about it on a regional basis, there's sort of three genuinely different problems here. Financing and politics in Europe, timing in India, and in China, a state building ahead of demand very, very deliberately.

Audrey Lee: Yeah, and underneath all three is the kind of the same problem if you like, that is capacity is rather limited. So, someone waits for it, and you also need someone to pay for it to solve the problem.

Daniel McCormack: Great, so I'd now like to shift gears, Audrey, and just talk a little bit about the age of grids, talk about missing grids, and what all of that means. In your paper, you talk about, both of these things, aging grids and missing grids as sort of two sides of the one problem. But let's take them one by one, and starting with aging, because I don't think people really have a feel for just how old this stuff is. Can you give the audience some numbers on how old the grids are?

Audrey Lee: Sure. For example, in Europe, 40% of distribution grids are already over 40 years old. And forecast out there says by 2030, we are going to reach more than 50%. And this isn't just 'things get worn out'. You have physically older transformers and circuit breakers that were engineered for one-directional power flow, and you have rooftop solar and EV chargers pushing power back into the same infrastructure. That is a genuinely different physical stress than it was designed for. So, it's a different kind of grid as well.

Daniel McCormack: You mentioned the Iberian blackout in the paper. Has anything like that happened since?

Audrey Lee: Unfortunately, yes, and it's starting to become a pattern, not just like, a one-off. In 2025 we saw blackouts in Chile and Mexico, and of course, the Iberian blackout you mentioned. And there was also a substation fire at Heathrow that shut down the airport, and you had an arson attack on grid infrastructure in Berlin. But what ties those incidents together is that aging, thinning-margin grid has very little redundancy left. A single point of failure, like, could be mechanical, could be weather-related or some deliberate attack, but all these problems will cascade further than it used to, because there's no spare capacity to absorb the shock anymore.

Daniel McCormack: Yeah, it sounds like the problems are only growing, and, you know, only going to grow further into the future if we don't take action. So, can you maybe give the audience a sort of, you know, a real-world example of what an aging grid like actually looks like and what the sort of consequences are of that aging?

Audrey Lee: Sure, and here's a very physical one. Transformer and high voltage circuit breaker lead times are now two to three years in Europe and North America, twice what they were back in 2019 pre-COVID. And that matters because you can approve a capex plan tomorrow, but you will still be waiting for years for the actual hardware.

Daniel McCormack: And, and let's turn to the sort of missing grid. Where is the grid just simply absent, not old?

Audrey Lee: In the emerging markets. For example, in South Africa, they are planning about 15,000 kilometres of new transmission line, and they need another one hundred and thirty three thousand GVA of new transformer capacity by 2034. That's a scale of what's missing in just one country, and if you scale it up to cover all other emerging markets, that's a huge number. And the gap isn't really about generation anymore as well. In most of these places, it's also about inspection, maintenance, tower resilience, and the specialist services that keep thin, aging, or even newly built lines actually standing through weather and load growth, that connects straight back to the infrastructure adjacency thesis from the earlier episode.

Daniel McCormack: You, we're talking about vegetation management, inspection, testing.

Audrey Lee: That's exactly that opportunity set, just relocated to grids that are either very old or badly built.

Daniel McCormack: So aging and missing grids aren't really two different problems. They're, they're the same underlying shortage of resilient capacity, but it's just showing up as sort of too old in Europe and not there yet in Africa and emerging markets, generally. Is that fair to say?

Audrey Lee: Yeah, I think so, and in both cases, someone has to wait. In Europe, that's 12,000 companies stuck in a Dutch connection queue. That's 640 million people without electricity access at all. That dollar figure is different by orders of magnitude, but the underlying scarcity is the same.

Daniel McCormack: Great. Let's switch to data centre demand, which is, you know, a fascinating area, and an area that's having a real impact on energy infrastructure and grids in particular. And perhaps let's start with something that's not talked about very much, which is regulation. Like, why is regulation relevant here, Audrey?

Audrey Lee: Yeah, the regulator matters here because everything downstream of this depends on it. For example, in June 2026, the FERC ordered all six US grid operators, so the PJM, MISO, SPP, CAISO, and a few others, to either justify or rewrite their tariffs for larger loads, like data centres, within 60 days. And the core question that FERC is wrestling with is exactly what we're talking about here. When a data centre needs billions of new transmission, does it pay the full cost, or does that cost get spread across everyone else's electricity bill?

Daniel McCormack: And how real is this cost-spreading risk, like for households and retail customers? You know, is this just theoretical, or, or is it already showing up in bills?

Audrey Lee: There are studies out there earlier this year that put the added cost from data centres at roughly $23 billion US dollars so far. Well, to be fair to the industry though, the picture is contested. So, some, for example, Senator Elizabeth Warren's claim that residents near data centres saw a bill rise 267%. When she said that, she was referring to the wholesale prices, not what shows up on a household bill. So, if we're thinking about average US residential prices, that's actually up around 40% over 5 years, and data centres are one driver among several, that you also have aging grids and equipment costs, among others that are driving up that bill. But the politically interesting bit is that you already see several hyperscalers pre-emptively pledge to cover their own costs. Microsoft made a broad pledge in January 2026. Several tech firms signed a non-binding ratepayer protection pledge at the White House in March as well.

Daniel McCormack: So, the industry is kind of seeing the backlash coming and is trying to get ahead of it.

Audrey Lee: Right, and the regulatory mechanism for that is the large load tariff. As of mid-2026, about 24 US states have approved at least one designed specifically to make sure large customers cover their own incremental costs rather than socialising them with, you know, average citizens.

Daniel McCormack: That's really interesting, and it's a rational kind of response from industry, to do that, right? But just turning to data centres, power supply, because data centres use a lot of power, they need power now. They sort of just can't wait the 5 to 8 years or however long it is to get a connection to the grid. So, Audrey, how are data centres solving this problem?

Audrey Lee: Yeah, Daniel, data centres are mainly turning to three different technologies. First of all, you have the most popular one, which is gas turbines, especially in the US. And it's so popular that it is now sold out. The three major manufacturers, including GE Vernova, that control roughly 75% of the market, they now all have lead times that stretch up to 8 years, and you see prices that are up as much as 300% in the past three years for gas turbines. In addition to gas turbines, other options out there include fuel cells, for example. That is the fast mover. Bloom Energy alone landed about 7.5 billion in data centre-related contracts in a single 90-day window in early 2026, including a big, like, multi-gigawatt deal with Oracle and its partnership with Brookfield expanded fivefold to 25 billion by the end of June 2026. Finally, another one that's being discussed quite frequently is nuclear and SMR, small modular reactors. That is more long-term though. As of May 2026, every major hyperscaler has signed at least one nuclear deal. But this would be a longer-term game as these nuclear SMRs take a long time to build.

Daniel McCormack: Got it. So, it's a bit of a three-way split here. We're talking, you know, turbines for scale, but with really long lead times. Fuel cells perhaps for speed. And then we've got nuclear, which is a bit more of a longer duration or longer term bet.

You said that, you know, the behind-the-meter workaround is mostly a US thing, because the grid there can't move fast enough, but, like, is anyone outside the US hitting, hitting the same wall?

Audrey Lee: Yes, indeed, and I would say that Southeast Asia is the clearest case here. It is more or less the same story. In the data centre hotspots, the grid is already tapped out. Like, for example, Johor is effectively exhausted, and you can't really fix that quickly. Augmenting a grid at scale runs like 5 to 10 years and even ordering a single large transformer is a multi-year wait because there's a global shortage.

Daniel McCormack: Thanks Audrey. Turning to batteries, like, where do batteries fit into all of this? I think, you know, people default to thinking of them as backup power. Is that the right way to think about them?

Audrey Lee: Well, it used to be, but not anymore. Traditional data centres use a 1 to 1 ratio of diesel generator for backup, but on newer projects, it drops to just 15 to 40% coverage. Why? Because now a battery is filling that gap. But I would say the more interesting shift is batteries becoming a queue-jumping tool as well, not just backup anymore. I'll give you an example here. In like December last year, a study modelling real data found that a traditional firm-only grid connection adds about 800 million in system costs per gigawatt of new data centre load. But if you pair it with a flexible grid connection with a 'bring your own capacity' kind of battery arrangement, that can eliminate that added cost and get about 500 megawatts of facilities online in about two years instead of 5 to 7. And there's also another live, named example. You have Portland General Electric work with developer Aligned and battery specialist Calibrant to connect a data centre using 30 megawatts or a 60-megawatt hour battery, rather than waiting for conventional infrastructure.

Daniel McCormack: Yeah, so batteries solve a different problem than turbines or fuel cells, and it's not base load, but, you know, speed and the ability to absorb the swings in load that we're seeing from AI.

Audrey Lee: Exactly. Turbines and fuel cells answer the question of how do I get firm power and then for battery, they answer, how do I get connected fast and how do I survive load swinging from 10 to 90% capacity in milliseconds. That's something no combustion turbine is built to do.

Daniel McCormack: Yeah, OK, so behind the meter, it's not one story. It's really three technologies, solving, you know, three different types of constraints. Like, really, really interesting. Thanks, Audrey. I'd now like to sort of turn to the final section of the podcast, which is a bit of a fun one, and that is, that is quickfire questions, if that's OK, Audrey. So, no long answers here, just short and sharp. Ten years from now, is the grid the hero of this story or the villain?

Audrey Lee: That's a tough one. I mean, hero, if the investment shows up on schedule. Villain, if it doesn't. You know, the, the AI story doesn't stop, it just relocates to wherever that has grid headroom.

Daniel McCormack: Fantastic. China and Europe, very different grid strategies. Which one would you bet on?

Audrey Lee: I mean, they are really different bets, like not really comparable. China for pace and scale, but for Europe is financing sophistication once the politics settle.

Daniel McCormack: Gas turbines, fuel cells, nuclear, we discussed all of these. Which, you know, which behind-the-meter bet ages best, do you think?

Audrey Lee: I would say batteries plus fuel cells for the next 5 years, nuclear for the 10-year view, once the first SMR fleet proves that they are indeed workable.

Daniel McCormack: Perhaps a cheeky one from me, but one thing in your paper you'd add if you were rewriting it today?

Audrey Lee: I would say it's the regional variation we just spent 15 minutes on. I'll give China its own section next time.

Daniel McCormack: Yeah, it makes sense. China's certainly, you know, a huge part of, a huge part of the story here. Well, thanks very much, Audrey. It's a very interesting topic, and, you know, it's one that's certainly not going away, given how old the grids are, given all of the pressure that is on them, and all of the investment that is very clearly needed. If there's one thing I think people should take away from today's podcast, it's that, you know, the grid isn't one story anymore. It's, you know, as many as 5 running on different clocks, funded by different people, and very much colliding in real time with the fastest moving demand shock, coming from data centres, that we've seen for decades.

So, Audrey, again, thank you.

For our listeners, please go and read the paper. It's fascinating. There's plenty of facts and details in there. You'll find a link in the show notes. We'll be back with more on this, you know, as the regulatory fight continues to play out.

Source: Ember, “World adds a record-breaking 814 GW of solar and wind in 2025”, March 2026.
Source: BNEF, “Grid Investment Outlook 2025”, December 2025.
Source: Eurelectric, EY and Imperial College London, “Grids for Speed”, May 2024.
Source: ENTSO-E, “28 April 2025 Blackout: Expert Panel Final Report”, March 2026.
Source: Baringa, “When the grid stops spinning: the inertia challenge in renewable energy”, September 2025.
See Macquarie Asset Management, Data centres: Powering the growth of AI and cloud computing (August 2025) for more detail.
Source: US Federal Energy Regulatory Commission, “2025 State of the Markets”, March 2026.
Electric Reliability Council of Texas. ERCOT is the operator of Texas’s electrical grid.
Source: BNEF, “Grid Investment Outlook 2025”, December 2025.
Source: Vercellino, R. et al., “Measurement of Generative AI Workload Power Profiles for Whole-Facility Data Center Infrastructure Planning”, National Laboratory of the Rockies, April 2026.
For example, The North American Electric Reliability Corp (NERC) issued a Level 3 alert (the highest level) in May 2026 following instances of data centres unexpectedly dropping load or oscillating demand rapidly, creating grid reliability concerns. Source: NERC, “Essential Action to Industry: Computational Load Modeling, Studies, Instrumentation, Commissioning, Operations, Protection, and Control Essential Action to Industry: Computational Load Modeling, Studies, Instrumentation, Commissioning, Operations, Protection, and Control,” May 2026.
Source: IEA, “Electricity 2026”, February 2026.
Source: IEA, “Building the Future Transmission Grid”, January 2025.
Source: IEA, “Breakthrough Agenda Report 2025”, October 2025.
Source: BNEF, “New Energy Outlook 2025: Grids”, August 2025.
Source: Lawrence Berkeley National Laboratory, “Interconnection Queue Dataset & Summarized Data Files, through 2025”, May 2026.
Source: European Commission, “European Grids Package” (COM/2025/1005), December 2025.
Source: European Parliamentary Research Service, “European Grids Package: Lessons Learnt from the Implementation of the TEN-E Framework”, December 2025.
Source: Eurelectric, EY and Imperial College London, “Grids for Speed”, May 2024.

IMPORTANT INFORMATION

The opinions expressed are those of the author(s) are as of the date indicated and may change based on market and other conditions. The accuracy of the content and its relevance to your client’s particular circumstances is not guaranteed.

This market commentary has been prepared for general informational purposes by the team, who are part of Macquarie Asset Management (MAM), the asset management business of Macquarie Group (Macquarie), and is not a product of the Macquarie Research Department. This market commentary reflects the views of the team and statements in it may differ from the

views of others in MAM or of other Macquarie divisions or groups, including Macquarie Research. This market commentary has not been prepared to comply with requirements designed to promote the independence of investment research and is accordingly not subject to any prohibition on dealing ahead of the dissemination of investment research.

Nothing in this market commentary shall be construed as a solicitation to buy or sell any security or other product, or to engage in or refrain from engaging in any transaction. Macquarie conducts a global full-service, integrated investment banking, asset management, and brokerage business. Macquarie may do, and seek to do, business with any of the companies covered in this market commentary. Macquarie has investment banking and other business relationships with a significant number of companies, which may include companies that are discussed in this commentary, and may have positions in financial instruments or other financial interests in the subject matter of this market commentary. As a result, investors should be aware that Macquarie may have a conflict of interest that could affect the objectivity of this market commentary. In preparing this market commentary, we did not take into account the investment objectives, financial situation or needs of any particular client. You should not make an investment decision on the basis of this market commentary. Before making an investment decision you need to consider, with or without the assistance of an adviser, whether the investment is appropriate in light of your particular investment needs, objectives and financial circumstances.

Macquarie salespeople, traders and other professionals may provide oral or written market commentary, analysis, trading strategies or research products to Macquarie’s clients that reflect opinions which are different from or contrary to the opinions expressed in this market commentary. Macquarie’s asset management business (including MAM), principal trading desks and investing businesses may make investment decisions that are inconsistent with the views expressed in this commentary. There are risks involved in investing. The price of securities and other financial products can and does fluctuate and an individual security or financial product may even become valueless. International investors are reminded of the additional risks inherent in international investments, such as currency fluctuations and international or local financial, market, economic, tax or regulatory conditions, which may adversely affect the value of the investment. This market commentary is based on information obtained from sources believed to be reliable, but we do not make any representation or warranty that it is accurate, complete or up to date. We accept no obligation to correct or update the information or opinions in this market commentary. Opinions, information, and data in this market commentary are as of the date indicated on the cover and subject to change without notice. No member of the Macquarie Group accepts any liability whatsoever for any direct, indirect, consequential or other loss arising from any use of this market commentary and/or further communication in relation to this market commentary. Some of the data in this market commentary may be sourced from information and materials published by government or industry bodies or agencies, however this market commentary is neither endorsed nor certified by any such bodies or agencies. This market commentary does not constitute legal, tax accounting or investment advice. Recipients should independently evaluate any specific investment in consultation with their legal, tax, accounting, and investment advisors. Past performance is not indicative of future results.

This market commentary may include forward looking statements, forecasts, estimates, projections, opinions and investment theses, which may be identified by the use of terminology such as “anticipate”, “believe”, “estimate”, “expect”, “intend”, “may”, “can”, “plan”, “will”, “would”, “should”, “seek”, “project”, “continue”, “target” and similar expressions.

No representation is made or will be made that any forward-looking statements will be achieved or will prove to be correct or that any assumptions on which such statements may be based are reasonable. A number of factors could cause actual future results and operations to vary materially and adversely from the forward-looking statements. Qualitative statements regarding political, regulatory, market and economic environments and opportunities are based on the team’s opinion, belief and judgment.

Other than Macquarie Bank Limited ABN 46 008 583 542 (“Macquarie Bank”), any Macquarie Group entity noted in this document is not an authorised deposit-taking institution for the purposes of the Banking Act 1959 (Commonwealth of Australia). The obligations of these other Macquarie Group entities do not represent deposits or other liabilities of Macquarie Bank. Macquarie Bank does not guarantee or otherwise provide assurance in respect of the obligations of these other Macquarie Group entities. In addition, if this document relates to an investment, (a) the investor is subject to investment risk including possible delays in repayment and loss of income and principal invested and (b) none of Macquarie Bank or any other Macquarie Group entity guarantees any particular rate of return on or the performance of the investment, nor do they guarantee repayment of capital in respect of the investment.

Past performance does not guarantee future results.

Diversification may not protect against market risk.

Market risk is the risk that all or a majority of the securities in a certain market – like the stock market or bond market – will decline in value because of factors such as adverse political or economic conditions, future expectations, investor confidence, or heavy institutional selling.

International investments entail risks including fluctuation in currency values, differences in accounting principles, or economic or political instability. Investing in emerging markets can be riskier than investing in established foreign markets due to increased volatility, lower trading volume, and higher risk of market closures. In many emerging markets, there is substantially less publicly available information and the available information may be incomplete or misleading. Legal claims are generally more difficult to pursue.

Currency risk is the risk that fluctuations in exchange rates between the US dollar and foreign currencies and between various foreign currencies may cause the value of an investment to decline. The market for some (or all) currencies may from time to time have low trading volume and become illiquid, which may prevent an investment from effecting positions or from promptly liquidating unfavourable positions in such markets, thus subjecting the investment to substantial losses.

Infrastructure companies may be subject to a variety of factors that may adversely affect their business or operations, including high interest costs, high leverage, economic slowdowns, surplus capacity, increased competition, commodity prices, regulatory and political developments, difficulty raising capital, and terrorist acts or political actions, and general changes in market sentiment.

The global financial crisis (GFC) refers to the period of extreme stress in global financial markets and banking systems between mid-2007 and early 2009.

Inflation is the rate at which the general level of prices for goods and services is rising, and, subsequently, purchasing power is falling. Central banks attempt to stop severe inflation, along with severe deflation, in an attempt to keep the excessive growth of prices to a minimum.

Stagflation occurs when persistent high inflation is combined with high unemployment and stagnant demand in a country’s economy.

Index performance returns do not reflect any management fees, transaction costs or expenses. Indices are unmanaged and one cannot invest directly in an index.

Macquarie Group, its employees and officers may act in different, potentially conflicting, roles in providing the financial services referred to in this document. The Macquarie Group entities may from time to time act as trustee, administrator, registrar, custodian, investment manager or investment advisor, representative or otherwise for a product or may be otherwise involved in or with, other products and clients which have similar investment objectives to those of the products described herein. Due to the conflicting nature of these roles, the interests of Macquarie Group may from time to time be inconsistent with the Interests of investors. Macquarie Group entities may receive remuneration as a result of acting in these roles.

Macquarie Group has conflict of interest policies which aim to manage conflicts of interest.

All third-party marks cited are the property of their respective owners.

Macquarie Asset management is a leading global asset manager offering a diverse range of investment solutions, including real assets, real estate, and credit.

This information is a general description of Macquarie Asset management only. The views expressed in this website represent those of the relevant investment team and are subject to change. No information set out above constitutes advice, an advertisement, an invitation, a confirmation, an offer or a solicitation, to buy or sell any security or other financial product or to engage in any investment activity, or an offer of any banking or financial service. Some products and/or services mentioned on this website may not be suitable for you and may not be available in all jurisdictions.

Investing involves risk including the possible loss of principal. The investment capabilities described in this website involve risks due, among other things, to the nature of the underlying investments. All examples herein are for illustrative purposes only and there can be no assurance that any particular investment objective will be realized or any investment strategy seeking to achieve such objective will be successful. Past performance is not a reliable indication of future performance.

Before acting on any information, you should consider the appropriateness of it having regard to your particular objectives, financial situation and needs and seek advice.

Other than Macquarie Bank Limited ABN 46 008 583 542 (“Macquarie Bank”), any Macquarie Group entity noted in this website is not an authorized deposit-taking institution for the purposes of the Banking Act 1959 (Commonwealth of Australia). The obligations of these other Macquarie Group entities do not represent deposits or other liabilities of Macquarie Bank. Macquarie Bank does not guarantee or otherwise provide assurance in respect of the obligations of these other Macquarie Group entities. In addition, if this website relates to an investment, (a) the investor is subject to investment risk including possible delays in repayment and loss of income and principal invested and (b) none of Macquarie Bank or any other Macquarie Group entity guarantees any particular rate of return on or the performance of the investment, nor do they guarantee repayment of capital in respect of the investment.

Additional important information (including regional disclosures)

[3246420]