Skip to content
Woyce Technologies
AboutTeamCareersContactStart a project →

Compute Diplomacy: How AI Datacenter Deals Became Foreign Policy

A look at how GPU exports, sovereign datacenter deals, and chip export controls have turned AI infrastructure into a tool of geopolitics.

Compute Diplomacy: How AI Datacenter Deals Became Foreign Policy — Woyce Technologies

A decade ago, if a country wanted to signal a strategic alliance, it might sign an arms deal or a trade pact. Today, it announces a gigawatt-scale datacenter. The currency of alignment has shifted from tanks and tariffs to GPUs and power grids, and the countries that get to buy the newest AI chips — and where they're allowed to run them — are being decided in the same rooms where trade agreements and defense pacts used to get hashed out.

This is compute diplomacy: the practice of using access to advanced AI computing power — chips, datacenters, cloud capacity — as a lever of foreign policy, and a close cousin of the domestic rulemaking we cover in compute governance. It sits at the intersection of export controls, sovereign wealth investment, energy politics, and great-power competition, and it's reshaping how governments think about AI the same way oil pipelines reshaped 20th-century geopolitics.

For anyone building or buying AI systems, this isn't abstract. Where chips can be shipped, which clouds can host which workloads, and how quickly capacity arrives in a given region increasingly depend on political decisions. This explainer covers what compute diplomacy means and how a typical deal is structured, why compute became a diplomatic asset in the first place, what it changes for businesses and technical teams, the open questions and limits of the current approach, and the signals worth watching next.

What compute diplomacy actually means

Compute diplomacy describes the set of state-level decisions that determine who gets access to advanced AI infrastructure — and under what conditions. It has three interlocking parts:

  • Export controls: Rules (mostly set by the US, given its dominance in high-end chip design) that restrict which countries and companies can buy advanced GPUs and the equipment used to make them.
  • Infrastructure deals: Bilateral or multilateral agreements to build large AI datacenters in specific countries, often bundling chip access with capital investment, energy commitments, and security guarantees.
  • Standards and alignment: Softer instruments — model-sharing agreements, cloud-access frameworks, and "trusted partner" designations — that determine which countries' AI ecosystems are allowed to plug into which supply chains.

None of this is entirely new. Technology has always had a security dimension — nuclear reactors, satellite launch capability, and encryption software were all treated as strategic assets long before AI. What's different with compute is the scale of capital involved (single datacenter campuses now cost tens of billions of dollars), the speed at which the underlying technology is advancing, and the fact that a huge share of the world's advanced AI chip supply runs through a small number of choke points — chip design (concentrated in the US), lithography equipment (concentrated in the Netherlands), and advanced fabrication (concentrated in Taiwan).

That concentration is what makes compute diplomacy possible. If any country could simply build its own advanced chips, there would be nothing to negotiate. Because almost none can, access becomes something that has to be granted, and grants become geopolitical instruments.

Advanced AI chip supply runs through three chokepoints, chip design in the US, lithography in the Netherlands and fabrication in Taiwan, which turns access into something governments grant.

The mechanics of a compute deal

A typical sovereign AI infrastructure deal now looks less like a trade contract and more like a mutual-defense arrangement with a balance sheet attached. The pieces usually include:

  1. Chip allocation: A negotiated cap or guarantee on how many advanced GPUs (or GPU-equivalents) a country's companies can import, often tied to end-use verification requirements.
  2. Capital and ownership structure: Joint ventures where a host country's sovereign wealth fund co-invests with a US hyperscaler or chipmaker, sharing in datacenter ownership rather than just hosting it.
  3. Energy commitments: Since a large AI campus can draw as much power as a mid-sized city, deals often include commitments to build out gas, nuclear, or renewable generation specifically to serve the site.
  4. Security and export conditions: Restrictions on re-export to third countries, physical security requirements, and sometimes personnel vetting for who can operate the facility.
  5. Political signaling: The announcement itself, timed to coincide with state visits or summits, functions as a public marker of alignment — a way for both governments to say "we are on the same side of this technology."

Anatomy of a sovereign AI compute deal: chip allocation, co-investment, and energy commitments on one side; export and security conditions plus political signaling on the other.

Why this matters right now

The clearest marker of how far this trend has gone is Stargate UAE, the joint US-Emirati AI infrastructure project whose first 200-megawatt phase is set to go live in 2026. It's described as the largest AI campus outside the United States built specifically to run on US-aligned technology — meaning US-origin chips, under terms negotiated between Washington and Abu Dhabi rather than left to the open market.

That framing matters. A 200-megawatt AI campus isn't a garden-variety cloud region — it's an amount of dedicated compute capacity that, a few years ago, would have been unusual even for a hyperscaler's home market. Building the largest one of its kind abroad, and doing so as a bilateral, government-brokered project rather than a private commercial expansion, is the clearest evidence yet that AI infrastructure siting decisions are now treated as foreign-policy acts, not just business ones — the same 200-megawatt-scale build-out we walk through in what actually happens inside an AI data center.

The logic driving deals like this runs in both directions:

  • For the country hosting the datacenter, it's a way to secure access to compute the country cannot yet build domestically, attract capital and jobs, and position itself as an AI hub — while implicitly picking a technology bloc to align with.
  • For the country whose chips power the site, it's a way to extend its technology standard into new markets, keep a strategically important partner inside its supply chain rather than a rival's, and gain some assurance about how the compute will be used.

This is precisely why chip export rules aren't purely economic anymore. A GPU sale is also a statement about which geopolitical camp a buyer is being permitted into, and datacenter siting has become one of the most visible ways that statement gets made.

What each side gets from a compute deal: the host country secures compute, capital and hub status, while the chip-supplying country extends its standard and keeps a partner in its supply chain.

Why AI compute became a diplomatic asset

Several forces converged to make this possible, and it's worth separating them because they explain why compute diplomacy is likely to persist rather than fade as a passing phase.

Supply chain concentration

Advanced AI chip production depends on a small set of irreplaceable inputs: leading-edge chip designs from a handful of firms, extreme ultraviolet lithography machines made by essentially one company, and advanced fabrication capacity concentrated in one region. This is a much narrower bottleneck than most other industrial supply chains, and narrow bottlenecks are exactly where states have historically found the most bargaining power — think oil chokepoints, rare earth minerals, or satellite launch capacity.

The scale of AI infrastructure investment

Frontier AI training and inference now require datacenter campuses that cost tens of billions of dollars and consume power on the scale of a small country. Very few private companies can finance that alone, which pulls sovereign wealth funds and state-backed investment vehicles into the picture — and once a government's balance sheet is involved, the deal inevitably becomes a matter of state interest, subject to the same scrutiny as any strategic asset purchase.

AI's dual-use character

Advanced AI models have applications that touch national security directly — cyber operations, intelligence analysis, military logistics, and increasingly, autonomous systems — which is part of why compute thresholds get treated as a proxy for risk in the first place. Governments that would never treat a commercial cloud contract as a security matter are treating AI compute access as exactly that, because the compute underlying a commercial chatbot and the compute underlying a defense-relevant AI system are, in practice, the same hardware.

Competition between blocs

The current environment features an explicit contest between US-aligned and China-aligned AI technology stacks, each trying to extend its standards, chips, and cloud platforms into third countries before the other does. Middle powers — Gulf states, Southeast Asian nations, parts of Europe — are being courted by both sides, which gives them unusual negotiating power precisely because their alignment isn't yet locked in.

That competitive dynamic is what turns ordinary commercial expansion into something governments feel compelled to manage directly. A hyperscaler opening a new cloud region used to be a routine business decision, weighed against local demand, tax incentives, and connectivity. When the same expansion also determines which technology ecosystem a country's banks, hospitals, and government agencies will build on for the next decade, it stops being routine. Both Washington and Beijing have effectively concluded that losing a country's AI stack to a rival bloc is a strategic setback worth countering with direct state involvement — financing, diplomatic pressure, or preferential export terms — rather than leaving the outcome to market forces alone.

Historical precedent

None of this is without precedent, which is useful context for judging how durable the current arrangement is likely to be. Civil nuclear technology followed a similar arc in the mid-20th century: a small number of states controlled reactor design and enriched fuel, and access was extended to allies through programs explicitly designed to bind recipient countries into a political orbit while denying the same technology to rivals. Satellite launch capability and, more recently, undersea cable infrastructure have followed comparable patterns — concentrated technical capability, treated as a strategic asset, extended selectively as a tool of alignment. Compute diplomacy fits this template closely enough that policymakers on both sides are consciously borrowing language and mechanisms from those earlier episodes, even as they insist AI is a fundamentally different kind of technology.

Benefits of Compute Diplomacy

Compute diplomacy has plenty of critics, and the limitations later in this piece are real. But the approach exists because it delivers something to each party involved, and understanding those benefits explains why it's spreading.

Host countries get capability they couldn't build alone

Almost no country can produce leading-edge AI chips, and building that capacity would take decades and enormous investment. A negotiated hosting deal gives a country large-scale compute within years rather than decades, along with capital, construction jobs, and a claim to regional AI hub status. For governments with ambitions in AI but no semiconductor industry, it's the fastest available route to meaningful capacity on home soil.

Chip-supplying countries extend their standards and keep oversight

For the country whose chips power a site, a deal keeps a strategic partner inside its technology ecosystem rather than a rival's. It also comes with conditions: end-use verification, re-export restrictions, and physical security requirements. Those conditions offer a degree of assurance about how dual-use compute is used that an open-market sale would not, which is the security case for managing access rather than either banning or freely selling it.

Businesses in host regions get local capacity

When a large AI campus is built in a region, companies there can run workloads closer to their users and data. That can mean lower latency for inference, simpler data residency compliance, and less dependence on capacity in distant jurisdictions. For regulated sectors like banking and healthcare in host countries, having compute available locally under clear terms can make AI projects feasible that would otherwise stall on data transfer questions.

Very large builds become financeable

Campuses costing tens of billions of dollars and drawing city-scale power are hard for any single company to fund. Bringing sovereign wealth funds into joint ventures spreads that capital burden, and government involvement can speed up energy and permitting decisions that would otherwise delay a project for years. Combining state capital with private technical capability is part of what allows these builds to happen at all.

Clearer rules for long-term planning

Despite the uncertainty, negotiated frameworks give infrastructure investors and cloud providers something to plan against: which countries can receive which chips, under which conditions. That is more predictable than ad hoc licensing decisions, even if the frameworks themselves can change. For the private sector, a known set of conditions is easier to work with than no framework at all.

Compute Diplomacy Use Cases

Compute diplomacy shows up through a recognisable set of instruments. These are the ways it's being applied today.

Government-brokered datacenter campuses

The most visible form is the sovereign AI campus built under a bilateral agreement. Stargate UAE is the clearest current example: a large AI site running on US-origin chips under terms negotiated between Washington and Abu Dhabi. The problem it addresses is the host country's lack of domestic chip capacity; the mechanism is a bundled package of chip access, investment, and conditions. Whether it delivers on both capability and alignment will be tested as the first phase goes live.

Tiered export licensing

Export rules are used to sort countries and companies by how freely they can buy advanced GPUs and chipmaking equipment. Licensing decisions, end-use checks, and caps on volumes let the exporting government extend access to partners while restricting rivals. The intended outcome is to slow the spread of frontier capability to adversaries without cutting off allied markets, though enforcement remains a weak point.

Sovereign wealth co-investment

Host-country investment funds increasingly co-own datacenters alongside foreign hyperscalers or chipmakers rather than simply hosting them. This gives the host a financial stake and some influence over the facility's direction, while the foreign partner gains capital and a committed long-term customer. The arrangement turns a commercial expansion into a shared strategic asset, with the political weight that implies.

Energy-for-compute arrangements

Because large AI sites need enormous power, some deals tie compute access to commitments to build generation capacity, whether gas, nuclear, or renewables. Countries with abundant energy or land use those resources as bargaining chips. The aim is to solve the power constraint at the same time as the chip constraint, though energy build-outs move far more slowly than diplomatic announcements.

Trusted-partner and cloud-access frameworks

Softer instruments, such as trusted-partner designations, model-sharing agreements, and cloud-access frameworks, determine which countries' ecosystems can plug into which supply chains and services. These shape where new model capabilities and cloud regions become available first. For businesses, they often show up indirectly as differences in what services are offered in which region.

What this means for businesses and builders

Compute diplomacy isn't just a story for foreign ministries. It has direct, practical consequences for any company that depends on AI infrastructure, and the effects are already showing up in procurement and planning decisions.

StakeholderPractical implication
Cloud/AI infrastructure buyersWhere you can deploy advanced-chip workloads is increasingly a function of your country's diplomatic alignment, not just your budget
Startups building on frontier modelsAccess to the newest model capabilities can lag by region depending on export rules and hosting restrictions
Enterprises with global operationsMultinational AI deployments may need to route different workloads through different jurisdictions to stay compliant
Governments without a chip industryDatacenter hosting deals become a primary lever for AI capability, made contingent on political alignment
Investors in AI infrastructureSovereign co-investment structures change the risk profile — geopolitical shifts can affect asset access, not just market demand

For companies, the practical upshot is that infrastructure planning now has a geopolitical dimension that didn't exist five years ago:

  • Data residency and compute residency are converging. It's no longer enough to know where your data lives; you increasingly need to know which chip generation, under which export license, is processing it.
  • Vendor selection carries political weight. Choosing a cloud provider with datacenters in a particular country can implicate a company in that country's alignment status, particularly for regulated industries.
  • Capacity planning has to account for policy risk. A change in export rules or a diplomatic rupture can constrain chip supply to a region with little warning, the same way sanctions can disrupt other supply chains.
  • Sovereign AI ambitions create new regional players. Countries building their own sovereign AI infrastructure through these deals are also building domestic AI ecosystems, which means new competitors, new talent pools, and new regulatory regimes for companies operating internationally.

None of this requires a company to be a defense contractor or a chipmaker. Any business running meaningful AI workloads abroad is now, whether it realizes it or not, operating inside a compute diplomacy framework set by governments it doesn't directly negotiate with.

For builders specifically — teams shipping products on top of frontier models rather than governments negotiating chip access — the practical questions are narrower but no less real. Which regions can reliably serve low-latency inference for a given model generation? Does a product roadmap assume access to compute that might not clear export review in a target market? Are there contractual dependencies on a single cloud provider whose regional footprint could shift if a hosting deal falls through or a diplomatic relationship sours? These aren't hypothetical planning exercises anymore; they're the kind of infrastructure due diligence that used to apply only to companies in explicitly regulated sectors like defense or telecommunications, and now applies to a much wider set of AI-dependent businesses.

The limitations and open questions

Compute diplomacy is a useful lens, but it's not a clean or fully settled system, and several structural problems remain unresolved.

Enforcement is hard. Export controls depend on end-use verification — confirming that chips sold to one country aren't quietly rerouted to another. Chips are small, valuable, and easy to move relative to the size of the deals involved, and verification regimes are still maturing. A rule on paper is only as good as the audits behind it.

The definition of "alignment" is fuzzy. Countries hosting US-aligned datacenters don't necessarily share US positions on every policy question, and hosting deals don't function like formal treaties with clear obligations. It's diplomacy by infrastructure, not by contract, which makes it flexible but also harder to hold either side accountable to.

Middle powers are hedging. Several countries actively courted for compute deals are simultaneously pursuing relationships with more than one technology bloc, treating chip access as a bargaining chip to extract better terms from all sides rather than committing fully to one camp. That's a rational strategy for the hosting country, but it undercuts the idea that these deals cleanly sort the world into stable blocs.

Energy is an underappreciated constraint. Every major compute diplomacy deal ultimately runs into the same wall: gigawatt-scale datacenters need gigawatt-scale power, and power infrastructure takes years to build regardless of how fast chip diplomacy moves. Political commitments can be announced quickly; power plants cannot.

The chip advantage may not last. Export-control-based advantage assumes the leading country retains a durable technology lead. If a rival closes the fabrication or design gap, the advantage embedded in these deals erodes, and the entire structure has to be renegotiated.

Common Compute Diplomacy Mistakes

The open questions above are structural. The mistakes below are what businesses and technical teams get wrong when they interpret or plan around compute diplomacy.

Treating announcements as available capacity

A headline about a gigawatt-scale campus is a statement of intent, not a cloud region you can deploy to next quarter. Chips have to clear export review, power has to be built, and facilities have to be commissioned. Teams that plan regional launches around announced capacity risk building roadmaps on infrastructure that arrives late, arrives smaller, or arrives with usage conditions they didn't anticipate.

Assuming today's export rules will hold

Export controls on advanced chips have been revised repeatedly, and deals that aren't formal treaties can shift with a change of government or a diplomatic rupture. Planning as if current rules are permanent leaves no room to respond when they move. Treating every arrangement as provisional, with a plan for what changes if access tightens, is more realistic than betting on stability.

Depending on a single region for critical workloads

Concentrating important AI workloads in one region, or with one provider whose regional footprint depends on a specific hosting deal, creates a single point of policy failure. If chip supply to that region is constrained or a facility faces new usage limits, there may be no quick fallback. The cost of designing for portability is usually much lower than the cost of an unplanned migration.

Equating data residency with full compliance

Many organisations already track where their data is stored. Fewer track which hardware, under which export license and which provider's conditions, is processing it. As compute residency and data residency converge, knowing only the data location can miss obligations attached to the compute itself, particularly for regulated industries and government-adjacent work.

Ignoring the political exposure of vendor choices

Choosing a cloud provider or region has always been a technical and commercial decision. Under compute diplomacy it can also signal alignment, or expose a company to the consequences of a host country's changing status. Regulated businesses that don't factor this into vendor selection can find themselves explaining a choice to regulators or customers after the fact.

Compute Diplomacy Best Practices

For businesses and builders whose AI depends on infrastructure they don't control, these practices reduce exposure to policy shifts.

  • Map your compute dependencies. Document where each AI workload runs, which provider and region it uses, what chip generation it depends on, and where its data lives. This is the baseline for every other decision, and most organisations discover dependencies they didn't know they had.
  • Design for regional portability. Keep critical workloads deployable in more than one region and, where practical, with more than one provider. Containerised deployment, infrastructure as code, and model choices that don't rely on a single region's hardware make a move feasible when you need it.
  • Write policy risk into contracts. Ask providers how they handle export-rule changes, regional capacity constraints, and facility restrictions, and seek terms covering notice periods, migration support, and service changes. Contracts that assume a stable environment leave you carrying the risk alone.
  • Track the signals that matter to you. Assign someone to follow export control updates, major hosting deals, and regional cloud announcements relevant to your markets. A short quarterly briefing is usually enough to avoid being surprised, as long as it ends with a clear note on whether anything changes for your own workloads.
  • Bring legal and compliance in early. For regulated industries or government-adjacent work, review compute location and vendor choices with legal and compliance teams before deployment, not after an audit question arrives.
  • Separate announced capacity from contracted capacity. In planning documents, distinguish capacity you can actually reserve today from capacity that has been announced, and build timelines only on the former.
  • Keep a tested fallback plan. For your most critical workloads, write down how you'd move them if access in one region tightened, and test the plan periodically so it isn't theoretical.

What to watch next

A few developments will indicate whether compute diplomacy hardens into a lasting feature of international relations or gets restructured as the underlying technology and politics shift.

  1. Whether more countries pursue "dual-track" deals — hosting infrastructure tied to more than one chip ecosystem simultaneously, which would signal that hedging is winning over exclusive alignment.
  2. How verification and end-use enforcement evolve, since the credibility of export-control-based diplomacy depends entirely on whether smuggling and diversion can be meaningfully policed.
  3. Whether energy constraints start to override chip constraints as the binding limit on where AI infrastructure can actually be built, regardless of political willingness.
  4. How sovereign wealth-backed AI investments perform financially, since a wave of underperforming compute joint ventures could cool government appetite for this model.
  5. Whether smaller or non-aligned countries build alternative compute coalitions among themselves rather than choosing a side, which would represent a genuinely new bloc structure rather than a binary one.

The Stargate UAE campus going live in 2026 will be an early test case for several of these questions at once — proof of concept for whether large, politically brokered AI infrastructure deals actually deliver the capability and alignment both sides are betting on, or whether the gap between diplomatic announcement and operational reality turns out to be wider than the press releases suggested.

Understanding how compute diplomacy shapes chip access and datacenter siting can help technical teams anticipate infrastructure risk earlier — and Woyce Technologies works with organizations navigating exactly these kinds of AI infrastructure and policy questions.

FAQ

What is compute diplomacy?

Compute diplomacy is the use of access to advanced AI computing infrastructure — chips, datacenters, and cloud capacity — as a tool of foreign policy, similar to how energy resources or arms sales have historically been used to build alliances and extend influence. In practice it shows up as export licensing decisions on advanced chips, large datacenter investments negotiated between governments and companies, and security conditions attached to who can use that capacity. The result is that access to frontier AI compute increasingly follows diplomatic relationships rather than purely commercial demand.

Why are AI chips subject to export controls?

Advanced AI chips are subject to export controls because production is concentrated in a small number of countries and companies, and because the resulting compute can be used for both commercial and national-security-relevant applications, making it a strategic asset in the same category as other dual-use technologies. In the US, these rules are administered by the Commerce Department's Bureau of Industry and Security, and they can cover the chips themselves, the equipment used to make them, and sometimes where and by whom the resulting compute can be used. The rules have been revised repeatedly, which is part of the uncertainty.

What is a sovereign AI datacenter deal?

It's an infrastructure agreement, usually involving a host country's sovereign wealth fund and a foreign chipmaker or cloud provider, to build a large-scale AI datacenter domestically. These deals typically bundle capital investment, chip access, energy commitments, and security conditions into a single negotiated package. For the host country, the appeal is domestic AI capacity, jobs, and a seat at the table. For the chip supplier and its government, the deal can extend influence and attach conditions on security, data handling, and who may use the facility. Terms are rarely fully public, which makes outside assessment difficult.

How does compute diplomacy affect businesses outside the tech industry?

Any business running significant AI workloads internationally can be affected through data and compute residency requirements, vendor selection constraints, and capacity availability that shifts based on the political alignment of the countries where their infrastructure is hosted. A retailer, bank, or manufacturer using cloud AI services may find that certain models or GPU types aren't offered in a preferred region, or that data must stay in-country. Building flexibility into vendor contracts and architecture reduces the cost of those shifts when they happen.

Is compute diplomacy the same as the US-China tech rivalry?

It overlaps with but isn't identical to that rivalry. Compute diplomacy also involves middle powers — Gulf states, parts of Europe, and Southeast Asian countries — negotiating access and alignment with multiple blocs, not just the two largest players. Many of these countries want access to leading-edge chips and cloud services while preserving trade relationships elsewhere, and they use investment capital, energy resources, and land as bargaining chips. That makes compute diplomacy a multi-sided negotiation, where alignment is often conditional, partial, and renegotiated as policy and technology change.

Can countries without advanced chip manufacturing still build AI capability?

Yes, primarily through hosting deals that bring foreign-made chips and datacenter investment onshore in exchange for political alignment and capital participation, rather than by developing domestic chip fabrication, which requires far longer timelines and larger investments. Other routes include renting capacity from foreign cloud providers, investing in AI talent and applications rather than frontier model training, and building smaller national models on available hardware. Each option trades some independence for speed, and each still depends on supply chains concentrated in a few countries.

What happens if export rules change after a datacenter deal is signed?

This is one of the open risks in the current system. Because these deals aren't formal treaties, a shift in export policy or diplomatic relations could constrain future chip supply to an existing facility, even though the infrastructure itself remains built and operational. A facility could end up running older hardware, face limits on who may use it, or need new licenses for upgrades. For companies relying on that capacity, the practical response is to avoid single-region dependence for critical workloads and keep a tested plan for moving them if access tightens.

Conclusion

AI capability now depends on physical infrastructure that very few countries and companies can produce: advanced chips, the equipment that makes them, and enormous amounts of power and datacenter capacity. That concentration is what turned compute into a diplomatic asset, traded through export licenses, sovereign investment, and conditional access in much the same way energy and arms have been.

The key insight for businesses is that compute availability is no longer only a pricing or procurement question. Which regions get new GPU capacity, which workloads can run where, and how quickly rules change all reflect political decisions that can shift with little warning.

There are real limits to what anyone can predict. Many deal terms aren't public, export rules have been revised repeatedly, and agreements that aren't formal treaties can change when governments or relationships do. Treat any current arrangement as provisional rather than settled.

A practical step is to map where your AI workloads and data actually run, which vendors and chip types they depend on, and what would happen if one region's access tightened. If you want help building that kind of resilience into your AI infrastructure, our technology consulting team can review your architecture with you.

WT

Woyce Technologies

AI & Engineering Team · Woyce

Woyce Technologies builds AI chatbots, LLM integrations, voice AI, and full-stack web applications for businesses in the US, UK, Europe & APAC. Based in Rajkot, Gujarat.

READY TO BUILD?

Let's build something
that actually works.

Tell us about your project. We'll be honest about whether we're the right fit — and if we are, we move fast.