Sector Intelligence

AI Is Becoming Strategic Infrastructure

Rgs 19, 2026

Ieva Bružaitė | Founder, Orion Capital | Published 19 September 2026,

Investment Thesis

AI is becoming strategic infrastructure in Europe and emerging as major investment theme. Artificial intelligence is moving from an application story into a strategic-infrastructure story. Europe’s security reset is accelerating demand not only for chips, but for the entire digital stack required to detect, interpret and respond to threats: accelerated compute, high-speed networking, sensor fusion, secure cloud and edge computing, cybersecurity, autonomous systems and the power infrastructure beneath them.

NATO’s direction is explicit. In July 2026, Allies announced more than $40 billion of investment in counter-drone capabilities over five years, alongside a plan to train five times as many drone operators by the end of 2027. NATO’s 2026 Alliance Digital Strategy calls for accelerated AI adoption, tactical-edge inference, resilient low-latency networks, data fusion from sensors and sensor-to-effector data flows.

The investment implication is broader than a conventional defence trade. The emerging opportunity is an AI resilience stack spanning semiconductors, networking, data platforms, cyber defence, autonomous systems and energy. The strongest beneficiaries may not all be labelled “defence companies”; many are already core suppliers to the civilian AI build-out.

The key risk is valuation – and it has grown more acute since this report was first published. A powerful structural trend does not guarantee attractive returns at every entry price, and the Federal Reserve’s 16 September rate hike, alongside ten-year Treasury yields near 5%, has raised the hurdle every long-duration asset must clear. Investors should separate technological importance from business quality, and business quality from the price paid.

ORION VIEW

Compute → Networks → Data → Cybersecurity → Autonomous Systems → Energy → Resilience

1. The Catalyst: Europe’s Security Architecture Is Becoming Digital

In the early hours of 15 September 2026, two Italian Eurofighter Typhoons operating under NATO’s Baltic Air Policing mission from Šiauliai shot down a drone that had entered Lithuanian airspace, downing it near the village of Pratkūnai in the Kaišiadorys district, roughly 100 kilometres west of Vilnius. Lithuania’s military said it was the first time a drone had been shot down in the country’s airspace; authorities declined to confirm the drone’s origin or purpose pending investigation, though the commander of Lithuania’s air force said the recovered wreckage suggested it was likely carrying explosives.

AI is becoming strategic infrastructure because modern defence increasingly depends on persistent sensing, rapid data processing, resilient networks and software-defined decision systems.

The incident should not be over-interpreted as an investment thesis by itself. Its significance is that it illustrates the operational problem Europe is already spending to solve: persistent surveillance, rapid identification, reliable communications and affordable interception.

Drone warfare has changed the economics of air defence. Relatively inexpensive unmanned systems can force defenders to employ costly sensors, aircraft or interceptors. The response therefore cannot be based only on adding more traditional missiles. It increasingly requires layered detection, electronic warfare, automated classification, inexpensive interceptors and software capable of combining multiple sensor feeds.

2. NATO Has Moved From Concept to Capital Allocation

At the NATO Summit Defence Industry Forum in Ankara on 7 July 2026, Allies announced that more than $40 billion would be invested in counter-drone capabilities over the next five years. NATO also said it would create a counter-drone marketplace for NATO-tested, interoperable systems and expand training so that five times as many drone operators are trained by the end of 2027, extending the multinational Flight Training Europe initiative to cover drone operators.

The wider digital transformation is even more important for investors. NATO’s Alliance Digital Strategy describes data as a strategic asset and calls for accelerated use of AI, including inference at the tactical edge. It envisages hybrid cloud connected to tactical edge computing through resilient, high-bandwidth, low-latency networks, as well as data fusion from sensors and integrated sensor-to-effector flows. In other words, the Alliance is describing a software-defined, data-centric defence architecture.

In this sense, AI is becoming strategic infrastructure across Europe’s security architecture rather then remaining a standalone software or application layer.

At the Ankara Summit, the Alliance also announced a €27 billion investment to modernise fuel storage and distribution infrastructure and support new facilities, including pipelines toward the eastern part of the Alliance – subsequently formalised when the North Atlantic Council approved the NATO Fuel Supply Chain Capability Programme Plan later in July. That is a reminder that digital resilience and physical resilience are converging: compute and software still depend on energy, logistics and hardened infrastructure.

3. The AI Resilience Stack

Compute: NVIDIA and AMD

The first layer remains accelerated computing. NVIDIA’s fiscal Q2 2027 revenue reached $96.2 billion, with Data Center revenue of $89.0 billion, up 117% year over year. The company cited demand from hyperscalers, AI-native companies, enterprises and sovereign customers. AMD is scaling from a smaller base: Q2 2026 Data Center revenue reached $6.7 billion, up 107% year over year, driven by EPYC processors and Instinct MI350 Series GPUs. The defence relevance is not that every military workload needs a hyperscale GPU cluster. It is that sovereign AI, simulation, computer vision, intelligence processing and edge inference expand the addressable market for accelerated compute beyond consumer-facing generative AI.

Networking: Broadcom and Arista

AI infrastructure is increasingly constrained by the ability to move data. Broadcom reported $16.7 billion of AI semiconductor revenue in fiscal Q3 2026, up 221% year over year and now 56% of total revenue, driven by custom AI accelerators and AI networking; management guided fourth-quarter AI semiconductor revenue to approximately $21.7 billion and disclosed customer order visibility extending to 2028. Arista reported Q2 2026 revenue of $3.036 billion, up 37.7% year over year, while introducing 1.6 Tbps AI fabric platforms. In civilian AI, networking connects accelerators and clusters. In defence, the same economic need appears in a different form: sensors, command systems, edge devices and decision-makers need resilient, low-latency connectivity.

Decision Intelligence: Palantir

The strategic value of data is realised only when it becomes a decision. Palantir’s Q2 2026 US government revenue grew 90% year over year to $809 million, while total revenue grew 93% to $1.94 billion. The company sits directly in the data-integration and decision-intelligence layer. This makes it one of the clearest public-market expressions of the convergence between AI, government workflows and operational decision-making – although its valuation can embed very high expectations.

Cybersecurity: CrowdStrike and Identity Security

Every additional connected sensor, cloud workload and autonomous platform expands the attack surface. Cybersecurity therefore becomes a structural companion to defence digitisation. CrowdStrike reported record Q2 FY2027 net new ARR of $333 million, up 51% year over year, and said Falcon Flex ending ARR exceeded $2.29 billion. The broader opportunity extends from endpoint protection and threat intelligence to identity, privileged access and zero-trust architecture. The strategic principle is simple: AI infrastructure cannot become mission-critical infrastructure unless it is trusted infrastructure.

Autonomous Systems and Counter-UAS

The most visible layer is the drone itself – and the systems designed to defeat it. NATO’s $40+ billion counter-drone commitment suggests a multi-year procurement cycle spanning surveillance, classification, jamming, interception, training and integration. The likely winners will include traditional defence primes, specialised private companies and technology suppliers embedded deeper in the stack. Investors should expect rapid technological turnover: the winning interceptor or sensor architecture of 2026 may not remain dominant in 2029.

Energy and Physical Infrastructure

AI is ultimately an electricity story. As AI is becoming strategic infrastructure, access to reliable power is becoming part of the competitive and security equation. Europe’s sovereign-compute and defence ambitions add to already enormous data-centre demand. In September, Google announced an investment of at least €13 billion (about $15.1 billion) in AI infrastructure in Finland across 2027 and 2028 – its largest single European investment – including a 22-year power purchase agreement with Fortum covering up to half the output of the Loviisa nuclear plant. That project is civilian, but it illustrates the same constraint facing strategic AI: secure compute requires abundant, reliable power. Grid equipment, generation, storage, cooling and hardened facilities are therefore second-order beneficiaries of the AI-resilience cycle.

4. The Discount Rate Just Moved

Four days after this analysis was first published, the single most important variable in the thesis changed. On 16 September the Federal Reserve raised its benchmark rate by a quarter point to a target range of 3.75-4.00% – the first increase since July 2023 – in a unanimous 12-0 vote. The updated projections were more consequential than the decision: 16 of 18 participants now see at least one further hike this year, with four penciling in two, and no cuts projected until 2028. Chair Kevin Warsh characterised the move as removing accommodation rather than braking the economy, while stating plainly that inflation has been too high.

AI is becoming strategic infrastructure, but infrastructure remains highly sensitive to the cost of capital.

The transmission to this thesis is direct. Ten-year Treasury yields have been trading close to 5%, and by Friday traders had increased bets on a further increase as soon as October. That matters because the AI resilience stack described above is, financially speaking, a very long-duration asset. Data centres, sovereign compute, grid connections, nuclear PPAs and multi-year defence procurement programmes all generate cash flows years into the future – and every one of them is now discounted at a materially higher rate than when NATO announced its counter-drone package in July. The strategic logic of the buildout has not weakened. The financial hurdle it must clear has risen.

Markets absorbed this with more composure than the headlines suggested. Semiconductors sold off sharply mid-week, then recovered: the sector index rose close to 3% on Thursday as yields stabilised below 5%, and finished the week down only about 1%. The Nasdaq ended higher while the Dow lost more than 1.5% – a divergence that suggests investors are repricing the cost of capital rather than abandoning the theme.

One further development deserves attention precisely because it comes from inside the industry rather than from the bond market. This week, leaders at Anthropic and OpenAI publicly called for a slowdown in frontier AI development on safety grounds, briefly rattling technology shares. For a thesis built on AI becoming strategic infrastructure, this is a genuine new variable: the defence and resilience applications described in this report – sensor fusion, edge inference, decision intelligence – depend far less on frontier-model capability than on deployment, integration and trust. It is possible for frontier development to slow while infrastructure adoption accelerates. But investors should not assume the two are entirely independent, and any regulatory response to such calls would land on the same suppliers.

5. The Public-Market Map

Company	Role in stack	2026 evidence	Primary investment risk
NVIDIA (NVDA)	Accelerated compute / sovereign AI	FY27 Q2 Data Center revenue $89.0B, +117% YoY	Expectations, export controls, customer concentration
AMD (AMD)	CPUs + AI accelerators	Q2 Data Center revenue $6.7B, +107% YoY	Execution versus NVIDIA; software ecosystem
Broadcom (AVGO)	Custom accelerators + AI networking	Q3 AI semiconductor revenue $16.7B, +221% YoY; Q4 guided to ~$21.7B	Large-customer concentration; cycle sensitivity
Arista (ANET)	High-speed AI networking	Q2 revenue $3.036B, +37.7% YoY	Hyperscaler concentration; valuation
Palantir (PLTR)	Data integration / decision intelligence	Q2 US government revenue $809M, +90% YoY	Very high expectations; procurement concentration
CrowdStrike (CRWD)	Cybersecurity / threat intelligence	Q2 FY27 net new ARR $333M, +51% YoY	Competition; platform execution; valuation

Note: The table maps strategic exposure; it is not a recommendation to buy or sell any security.

6. Why Europe Could Become the Next Catalyst

Europe spent much of the previous decade optimising for efficiency. The emerging regime prioritises resilience. Defence budgets are rising, procurement is accelerating, sovereign technology has become a policy objective and the eastern flank requires both physical and digital infrastructure. This is not merely a defence-spending story. It is a reindustrialisation story.

AI is becoming strategic infrastructure as Europe links security, sovereign technology and industrial capacity.

The most powerful investment cycles occur when several capital-expenditure waves overlap. Europe is simultaneously trying to expand defence production, harden energy systems, build sovereign AI capacity, strengthen cybersecurity, modernise networks and reduce strategic dependencies. Those objectives share suppliers, power requirements and digital infrastructure.

This is why the second-order AI trade may prove more durable than a narrow “chatbot” narrative. The same GPU, switch, optical component, security platform or data layer can serve commercial AI today and strategic infrastructure tomorrow. Dual-use economics can enlarge markets without requiring investors to predict a single defence programme.

7. What Could Break the Thesis?

Valuation compression. Structural winners can still be poor investments when growth expectations are priced too aggressively.
Procurement friction. European defence purchasing remains fragmented and government sales cycles can be slow – as the summer’s delay over the fuel-infrastructure programme illustrated.
Technology turnover. Drone and counter-drone architectures are evolving rapidly; today’s leader can be bypassed.
Export controls and geopolitics. Semiconductor restrictions can alter addressable markets and supply chains.
Customer concentration. Several AI-infrastructure suppliers depend heavily on a small number of hyperscale customers.
Budget substitution. Higher defence budgets do not guarantee that spending flows to listed technology companies rather than domestic or private suppliers.
Cyber and operational failures. As infrastructure becomes software-defined, a single security or reliability failure can carry outsized reputational and financial consequences.
The cost of capital. With the Fed hiking and ten-year yields near 5%, every long-duration AI and infrastructure project faces a higher hurdle rate – a risk that applies regardless of how strong the underlying demand proves to be.
Self-imposed industry constraints. Calls from within the AI industry itself to slow frontier development, and any regulation that follows, could alter both sentiment and deployment timelines across the stack.

8. What Orion Capital Is Watching Next

Rather than treating every security incident as a trading signal, the better approach is to monitor capital allocation. If AI is becoming strategic infrastructure, capital allocation becomes one of the clearest signals of where the opportunity is moving. The thesis strengthens when multiple independent indicators move together: NATO and national procurement, sovereign-AI projects, edge-compute deployments, networking orders, cybersecurity budgets, power contracts and defence-oriented data-platform adoption. To that list we would now add one macro variable: the path of long-term interest rates, which determines the price investors are willing to pay for all of it.

Signal	What would confirm the thesis
Counter-UAS procurement	Does the $40+ billion NATO commitment translate into recurring orders and standardised architectures?
Sovereign AI	Do European governments build domestic compute capacity and secure clouds at scale?
Tactical edge	Does AI inference move from central data centres into deployed systems and sensors?
Networking	Does low-latency, high-bandwidth connectivity become a defence bottleneck similar to AI data centres?
Cybersecurity	Does zero-trust and identity security become mandatory across defence digitalisation?
Power	Do data-centre and sovereign-compute projects lock in nuclear, grid and storage capacity through long-term contracts?

Conclusion: From Intelligence to Resilience

Every major technology eventually disappears into infrastructure. Electricity moved from novelty to industrial foundation. The internet moved from websites into banking, commerce, government and communications. AI is becoming strategic infrastructure, beginning the same transition from visible technology to an underlying economic foundation.

The chatbot made artificial intelligence visible. Strategic infrastructure may make it indispensable.

For Europe, the transition is arriving precisely when security, sovereignty, energy and technological capability are becoming inseparable. For investors, that expands the map far beyond one chipmaker or one defence contractor. The emerging stack links compute, networks, data, cybersecurity, autonomous systems and power.

The first AI wave was about intelligence. The next may be about resilience.

Ad Astra.

Disclaimer: This publication is for informational and educational purposes only and does not constitute investment advice, an offer, solicitation, or recommendation to buy or sell any security. Market conditions, company fundamentals and valuations can change rapidly. Readers should conduct their own research and consider their financial circumstances and risk tolerance.

Sources and Notes

About the Author

Ieva Bružaitė is the Founder of Orion Capital, an independent research platform covering AI, semiconductors and long-term investing.

← Back to Orion Capital