Key Points
- The Japanese government’s new public–private investment roadmap for 17 designated strategic sectors targets ¥101.6 trillion (~$635 billion) in investments in AI and semiconductors. By comparison, the target for investment related to power generation is only ¥15.2 trillion (~$95 billion). The two figures cannot be compared directly but nonetheless beg the question: can Japan secure enough of the electricity that AI data centers and semiconductor plants need to run?
- The problem goes beyond the “quantity” of electricity. Both frontier sectors such as AI data centers and semiconductor plants as well as the materials and components industries in which Japan is already a market leader all require large volumes of high-quality power. If AI and semiconductor investments lead to a tightening of the supply–demand balance for electricity, it could worsen the competitive conditions for the materials and components industries.
- Behind the failure to narrow down the 17 sectors lies not a problem of vested interests but one of institutional design: a budget-request process with no ceiling. Unlike a private investor, the government has no single actor to bear the pain of cutting its losses when an investment fails to deliver requisite returns. Precisely for that reason, making the proposed “flexible review” process effective requires building deadlines and criteria for scaling back or withdrawing into the system in advance.
Dollar equivalents are approximate, converted at ¥160/$ as of August 2026, and will shift with the exchange rate.
Are We Cutting Into Our Own Winning Hand?
In the Japanese government’s public–private investment roadmap for its “17 strategic sectors,” the figure of ¥101.6 trillion for AI and semiconductors appears, the largest for any single sector. AI data centers and semiconductor plants require enormous amounts of electricity. Yet if one pulls out the investment directly related to power generation from the same roadmap, it is only ¥15.2 trillion. The two figures cannot be compared directly. Even so, the question remains: while promoting AI and semiconductor investment on this scale, can Japan secure enough of the electricity it needs?
I do not see this as a “skewed allocation”—thick on the sectors where Japan is losing, thin on those where it is winning. There is a more structural problem at work here. If large-scale investment in AI and semiconductors pushes up power demand, it can worsen competitive conditions for the industries in which Japan is already strong. In effect, the thick side eats the thin side.
More specifically, if the quantity and quality of available power does not keep pace with the expansion of demand from new sectors, the result will be competition for electricity that leads to higher electricity prices and strain on the stability of supply. Those forced to bear the burden will be the very sectors where Japan currently holds high international competitiveness, such as magnets, semiconductor materials, and ultra-high-purity chemicals. Like AI data centers and semiconductor facilities, the manufacture of these products requires large volumes of high-quality power and, in some processes, even a momentary voltage dip can lead to defective products.
In other words, if things proceed as they are, Japan risks failing to catch up sufficiently in the frontier fields it is targeting while also damaging its existing competitive advantages.[DP1]
This essay examines the investment allocation across the Takaichi government’s 17 strategic sectors through the shared constraint of “electricity.” Why did the allocation turn out this way? And where should it be changed?
What Cannot Be Discarded Is Not the Fault of Vested Interests
Strategy is deciding what to discard. This is common sense in management. Apply that yardstick, and the government’s selection of 17 sectors and 62 items looks like an everything-for-everyone list. So why couldn’t it be narrowed down? Many attribute the cause to the political distribution of benefits but I think an institutional factor has been overlooked. If one were to explain the issue as a problem of distributing benefits, the prescription becomes political reform. If institutional design is also a cause, what is needed is the far more mundane task of redesigning the budget process.
According to government documents, the budgeting mechanism for the newly created investment categories will allow necessary funding amounts to be requested as appropriate without setting a request ceiling. In this case, the term “request ceiling” refers to the upper limit typically imposed in advance on each ministry’s budget requests. The act of “discarding” and choosing between priorities occurs only when the source of funds is finite. Absent a ceiling, there is little rational reason for the requesting side to deliberately pare down its request. My reading is that the number 17 is less the product of political maneuvering than a natural consequence of a request process with no ceiling.
In fairness to the government, it has said it will flexibly review budgets whose investment-inducing effects are weak. This is the right approach. The problem is that this “flexible review” has not yet been translated into concrete criteria—which indicator, by when, and at what level—that would trigger a program being scaled back or abandoned. What deserves criticism is not the government’s intent but this gap in institutional design.
Given this, how should the budget be reorganized? The criterion should not be “whether Japan is winning.” It is how far an additional ¥1 trillion of investment can change the outcome. The winner in cutting-edge logic semiconductors, for example, is decided by factors such as the availability and scale of manufacturing equipment and by the customer base. There is thus a limit to how far Japan’s fiscal support alone can move it. By contrast, the quantity and quality of electricity, and the supply capacity of materials, move more clearly in response to policy levers.
Electricity Is Not a Sector. It Is the Denominator.
As noted at the outset, the largest investment target across the 17 designated sectors is ¥101.6 trillion for AI and semiconductors. By contrast, the target for the resources, energy security, and GX field, which includes power generation, comes to only ¥28.8 trillion. That is below the ¥32.7 trillion put forward for the single item “cloud, data centers, and storage batteries” even before considering that not all of the ¥28.8 trillion goes to power generation.
If we break down that ¥28.8 trillion figure into its subcomponents, it includes ¥4.2 trillion for green steel and ¥3.2 trillion for GX chemicals—demand-side decarbonization investments—and ¥6.2 trillion earmarked for hydrogen, which is likewise not limited to power generation. Investments directly related to power generation are offshore wind (¥5.1 trillion), next-generation innovative reactors (¥5.0 trillion), next-generation solar cells (¥4.1 trillion), and next-generation geothermal (¥1.0 trillion), which add up to a total of ¥15.2 trillion.
Of course, the ¥101.6 trillion for AI and semiconductors and the ¥15.2 trillion for power generation differ in both their targets and their nature. One cannot simply compare them and conclude that power investment is “insufficient.” However, the government itself anticipates that power demand may be set to rise sharply. The power-demand outlook for fiscal 2040 is 0.9 to 1.1 trillion kWh; at the high end, that is an increase of roughly 200 billion kWh a year from about 0.9 trillion kWh in fiscal 2022. Presuming large-scale investments in AI data centers and semiconductor plants go forward, the question of whether the electricity supply to support them can be secured is unavoidable.
Here, though, the issue is not the quantity of electricity alone. It is also vital that the necessary quality of power be available in the required amounts.
The Question Is Not Quantity Alone—Four Conditions for Electric Power Supply
Generation volume alone does not capture the conditions that shape a company’s locational competitiveness. Here I divide them into four. First: does the power stay on? This refers to the frequency and duration of outages as well as the speed of recovery. Second: does it hold steady? This refers to both the size of “instantaneous voltage drops,” such as a momentary dip in voltage from a lightning strike, and of voltage and frequency fluctuations. Third: can you connect? This refers to whether a large consumer can connect to the grid at the required time and with the required capacity. Fourth: at what price can you use it? This refers to the unit price of electricity, grid usage fees, and the cost of balancing capacity.
Operating an AI data center requires that all four conditions be met. Large-scale AI processing has sharp load swings, with power demand fluctuating greatly over short intervals, which tests not only average power consumption but how far the power system can accommodate instantaneous change. Even if annual generation volume balances out on paper, it means nothing if power of the required capacity and quality cannot be received in the required place and at the required time.
Of the government’s proposed ¥15.2 trillion in power-generation-related investment, offshore wind and next-generation solar cells account for ¥9.2 trillion, or about 60 percent. The problem is not the size of the investment in variable power sources. It is the extent to which these investments are supported by corresponding investments in storage batteries, synchronous condensers (rotating machines that stabilize voltage and frequency), grid inverters, balancing capacity, the backbone transmission grid, and long-term backup power. In other words, the question we should ask is whether investment in the equipment that underpins stable power supply is positioned sufficiently and explicitly within the same roadmap.
Japan’s Advantage in High-Quality Power
There is something else being overlooked here. Excluding years of major natural disasters, Japan’s annual power-outage duration averages about 20 minutes, among the shortest in the world. In major regions of the United States, years where the comparable figure exceeds 100 minutes are not uncommon. The Japanese government and the Organization for Cross-regional Coordination of Transmission Operators (OCCTO) publish this outage frequency and duration as an indicator of Japan’s power quality.
This high level of performance is the accumulation of decades of capital investment and operational know-how and is not something that other countries can imitate in a short time. In the competition to site data centers, Japan may not win on land costs or electricity prices but this is precisely why there is room to leverage the fact that Japan’s power does not stop or waver as a locational competitive strength.
However, this point comes with two important caveats. First, the published indicator is skewed toward measuring the first key condition for high-quality power outlined above—that is “does it stay on” (i.e., outage duration and the like). No comparably clear metrics are published for evaluating instantaneous voltage drops, frequency stability, and the connectivity available to large consumers. This is important given that what matters for AI data centers and the materials and components industries is not only the scarcity of outages but also these issues of power quality and connectivity.
Second, quality is not maintained if left alone. As the share of variable power sources rises and conventional rotating machines such as synchronous generators decline, additional measures are needed to stabilize frequency and voltage. Today’s “about 20 minutes” of annual outage time is a legacy of past investment not a guarantee of future performance.
Investment in Losing Sectors Erodes the Moat of the Winning Ones
The issue here is that the manufacturing industries in which Japanese firms already hold high competitiveness are competitors for the same quality power that frontier sectors require.
Japan’s advantage lies not in finished products but in specific manufacturing processes and in certain material and component segments, areas where Japanese firms already hold high international market share. For example, with regard to permanent magnets, the growth strategy roadmap itself points out that Japan is effectively the only country (outside a “certain country”) with the capacity to produce certain high-performance magnets.
The same is true in semiconductors. Japan has fallen far behind in cutting-edge logic semiconductors, but the Ministry of Economy, Trade and Industry’s “Semiconductor and Digital Industry Strategy” notes that Japanese firms hold about a 30 percent share of the market for manufacturing equipment—second only to the United States—and about half the share of key materials and components, making them indispensable to the overall supply chain. (The latest version of that strategy also points out that Japan’s market share is shrinking in some equipment areas. Advantage is not fixed.)
The competitiveness of Japan’s materials and components industries cannot be explained by power quality alone. The accumulation of technology and relationships with customers are also key factors. However, stable, high-quality power is a foundational condition for the success of these industries, where voltage fluctuations or instantaneous outages can lead to a halted production line or defective products. This presents a dilemma for firms involved in sectors such as ultra-high-purity chemicals, semiconductor materials, or precision processing. High-quality power is an essential input to their business but not something that individual firms can secure through their own efforts alone.
As noted above, AI data centers and the materials and components industries are thus potential competitors for the same critical input: high-quality power. If large-scale investment in the AI and semiconductor fields spurs the siting of data centers and semiconductor plants, power demand rises accordingly. If power supply and grid development cannot keep pace with demand, the negative effects may impact existing manufacturing through higher electricity prices and strain on stable supply. The structure described at the outset of this essay operates through the concrete circuit of electricity.
To be fair, Japanese policymakers have not ignored the materials field. The aforementioned roadmap sets targets such as an 80 percent global share in 2030 for high-functional metals used in semiconductor circuits. This suggests a further question of why a target of “80 percent” can be set for certain materials, while for semiconductors “recovery of lost ground” has not been translated into a comparable market-share target? Behind the ability to set a concrete numerical target of 80 percent must lie a certain confidence in the existing technological base. Conversely, the very way targets are set seems to mirror the government’s degree of conviction.
What Must Be Protected Are the Conditions That Underpin Competitive Advantage
Let me answer an anticipated objection. Since materials are highly profitable and private investment is already flowing in, is it not rational to allocate public funds to the frontier fields where private capital is reluctant to enter, namely AI and semiconductors?
There is something to this point, hence why I do not argue for “more subsidies for materials.” What underpins Japan’s hard-to-see competitive advantage is not subsidies themselves but high-quality power, appropriate export controls, and a deep layer of engineers. None of these can be sufficiently secured by corporate effort alone, and that is precisely why policy levers work.
Investment in electricity is also an investment in the chokepoint industries such as the materials and components industries in which Japan is already strong. Energy and manufacturing are thus not two separate proposals. They are one strategy that should be designed as an integrated whole.
Three Things to Change
Given the above, investment in the government’s 17 designated sectors needs to be designed taking into account not only allocations to individual sectors but also the shared constraints and the rules for review. Specifically, the following three points should be revised.
First, elevate electricity and treat it not as a “sector” but as the “common denominator” beneath every sector—the shared base on which all of them run. The government should seek to build up this common denominator while evaluating and handling quantity and quality as distinct, measurable targets. This should include re-examining the scale of power-generation-related investment and, within the same roadmap, mapping the expansion of variable power sources against the buildout of storage batteries, synchronous condensers, balancing capacity, and the backbone transmission grid. For the backbone grid, the master plan for the wide-area interconnection grid already outlines the scale of long-term development and investment in such transmission networks should also be explicitly positioned as strategic investment.
Second, manage power quality as a national competitive asset. This should include positioning not only outage duration but also the frequency of instantaneous voltage drops, frequency stability, and the connection waiting period for large consumers as indicators that shape the siting of AI data centers and the competitiveness of the materials and components industries—and then setting target values for these indicators. Japan cannot win by competing on price; it should look to markets that value (and pay for) quality.
Third, translate “flexible review” in the budget process into more concrete criteria. For each item, put in writing, at the point of commitment, which indicator must reach what level by when, failing which the commitment to a project is scaled back or withdrawn. The level at which to cut losses and abandon a project can only be set before making the initial investment. Once the investment is made, judgment inevitably softens and the temptation to throw good money after bad increases.
Finally, let me restate the argument of this essay. The problem the government’s current strategy faces is not investing in fields where Japan is less competitive. It is that such investment risks cutting into the competitive advantage Japan already holds.
Magnets, semiconductor materials, and ultra-high-purity chemicals cannot be protected by subsidies alone. What protects them is power that does not stop and does not waver, export controls, and the layer of engineers who carry the front line. None of these can be sufficiently secured by corporate effort alone. That is precisely why there is a critical role for government.
The figure of ¥370 trillion (~$2.3 trillion) should be applauded as a statement of resolve. But considered as an investment portfolio, the overall design is not yet sufficiently developed. An investment category with no exit criteria risks becoming a vessel for budget requests rather than an investment strategy. That withdrawal is hard for the government to decide is not a matter of politicians’ character. It is because, unlike a private investor, the government has no single actor to directly bear the “pain of cutting losses.” That is precisely why this weakness must be compensated for through institutional design. Unless the design goes that far, the next administration will end up drawing up an 18-sector strategic investment plan.
*The views expressed in this article are solely those of the author and do not represent the views of any organization, including the institution with which the author is affiliated.
Source: “Japan Growth Strategy Meeting” (Official Website of the Prime Minister’s Office of Japan), edited by DCER
