祝融说。

A magnificent building rises from a solid foundation. A powerful empire is built upon a fertile land. Likewise, a national intelligent computing center -- this "super factory" and "decision-making hub" of the digital age -- has its success and value determined, to a large extent, by its physical "foothold" in the world.

Site selection and energy consumption, seemingly technical matters belonging to the early preparatory stages of a project, are often reduced in many traditional infrastructure projects to a mere accounting exercise: Where is the land cheapest? Where is the electricity cheapest? Where are policy subsidies the greatest? Completing the journey from planning to the flipping of the switch by the shortest path and at the lowest cost seems to be the project manager's greatest achievement.

Yet for an intelligent computing center, this short-sighted, purely "accounting" mindset is extremely dangerous and misleading, because it ignores a physical fact that is at once the most basic, the most plain, and the most profound. This fact is the logical starting point for understanding every energy issue of the intelligent computing center, and even the "Eastern Data, Western Computing" national strategy itself.

That fact is: computing power is heat.

1.1 Computing Power Is Heat: The Physical Nature of the Intelligent Computing Center and the National Chessboard

We are often deceived by the virtuality of the digital world. Code, data, models... these concepts feel light and ethereal, as if they existed in a "cloud" parallel to our physical world, bearing no mass and consuming no energy. We speak of "going to the cloud," as though data truly rises lightly into the sky like water vapor.

This is a beautiful misunderstanding.

Every operation in the digital world must leave a trace in the physical world. That trace is heat.

Let us peel back the layered racks, cables, and software of an intelligent computing center and look straight at its most essential unit -- the chip. Whether CPU, GPU, or dedicated ASIC, a chip is at bottom a super switch composed of billions or tens of billions of tiny transistors. Every calculation, whether a simple addition or a complex matrix multiplication, corresponds to billions of "on" and "off" state transitions performed by these transistors at astonishing speed. According to the fundamental laws of physics, any macroscopic irreversible process leads to an increase in entropy; and at the microscopic level, every change in a transistor's state, no matter how efficient, inevitably entails an expenditure of energy. This dissipated energy is ultimately released almost entirely in the form of heat.

The heat produced by a single transistor is negligible. But when tens of thousands of high-performance chips roar simultaneously at near-100% load inside an enclosed space, the heat generated is staggering. A high-performance server equipped with eight top-tier GPUs typically draws on the order of 10 kilowatts at full load (a back-of-envelope estimate assuming per-card power in the hundreds of watts up to a kilowatt) -- roughly comparable to the heating power of dozens of household air conditioners. And a large intelligent computing center typically deploys thousands, even tens of thousands, of such servers.

We must therefore establish a new understanding: a national intelligent computing center is, in its physical essence, not an information "library" but an energy "conversion furnace." It draws in vast amounts of electrical energy and, through the "smelting" process of computation, forges raw, chaotic data (high information entropy) into structured, orderly knowledge and decisions (low information entropy). In doing so, it inevitably produces a byproduct -- a vast quantity of heat (high thermodynamic entropy) that must be expelled.

1.1.1 The Entropy Paradox: The Duality of Information Entropy Reduction and Physical Entropy Increase

Here we encounter the first core paradox of the intelligent computing center, and the key to understanding the heart of this book.

From the perspective of information theory, the intelligent computing center is a powerful "entropy reducer." It opposes the "entropy of information overload," the "entropy of decision-making complexity," and the "entropy of future unpredictability" described in the introduction. Through its powerful convergence capability, it injects order and certainty into a chaotic world. This is its mission and its value.

From the perspective of thermodynamics, however, the intelligent computing center is an out-and-out "entropy increaser." It follows the second law of thermodynamics, converting high-quality, orderly electrical energy into low-quality, disorderly heat, discharging it into the environment, and thereby increasing the total entropy of the entire physical system.

This duality of "information entropy reduction" and "physical entropy increase" is the fundamental contradiction that every manager of an intelligent computing center must confront. Our task is to buy the greatest "information entropy reduction" at the lowest cost of "physical entropy increase." And the first -- and most critical -- step toward resolving this contradiction is site selection.

At its core, site selection answers a single question: where on the national territory should we place this enormous "heat source" and "power hog" so that its negative impact on the physical environment is minimized while the positive information value it generates is maximized?

If we were to follow the traditional commercial logic of "be close to the market" and pile every large intelligent computing center into the first-tier coastal cities of the east, we would quickly face a disaster. Beijing, Shanghai, and Shenzhen are densely populated, with land as precious as gold and severely constrained energy quotas. Building a 100,000-server intelligent computing center in such a city would be like erecting another steel mill on the scale of Shougang beside Wangfujing or Nanjing Road in the city center. It would not only impose an unbearable load on the power grid; the heat it dissipates would intensify the urban heat island effect, and the vast water its cooling systems require would compete with residential life and industrial production for already scarce quotas.

Such a layout would infinitely amplify the "physical entropy increase" of the intelligent computing center, and its costs and negative externalities would soon exceed the benefits of its "information entropy reduction." It is an unsustainable model of development.

It is precisely against this backdrop that "Eastern Data, Western Computing" -- a national strategy of far-reaching vision -- came into being.

1.1.2 "Eastern Data, Western Computing": From an IT Layout to a National Thermodynamic Solution

If one understands the "Eastern Data, Western Computing" project merely as "taking the East's data to be computed in the West," one greatly underestimates the profound strategic thinking behind it. It is not a simple scheme for remotely deploying IT resources. It is a systematic, thermodynamics-based solution, devised at the national level, to the "entropy paradox" of intelligent computing centers.

It is a re-examination and optimized allocation of the nation's overall resource endowments, an ambition and boldness comparable to the digital era's "South-to-North Water Diversion" and "West-to-East Gas Transmission."

• "Eastern Data" -- the source of demand, the outlet for value

  The eastern region, especially the Beijing-Tianjin-Hebei region, the Yangtze River Delta, and the Guangdong-Hong Kong-Macao Greater Bay Area, is the most economically vibrant and digitally advanced part of the country. It concentrates the greatest number of financial institutions, internet companies, high-end manufacturers, and research institutes. These are the largest producers of data and the most pressing consumers of computing power. Urban governance, financial transactions, autonomous driving, the industrial internet... these application scenarios constitute the basic foundation of "Eastern Data."

  At the same time, the eastern region is where the value of computing power is ultimately realized. The decisions generated by computing power must land here; the industries incubated by computing power must take root here. The East, therefore, needs the "interfaces" and "showcases" of computing power, not its "production workshops."

• "Western Computing" -- the base for energy, the fertile ground for heat dissipation

  By contrast, China's western region possesses natural advantages for building large and ultra-large intelligent computing centers -- advantages that answer precisely to the physical nature of such centers as "heat sources" and "power hogs."

  1. Energy endowments (solving the "power hog" problem): The western region is the nation's strategic energy rear base. The wind of Xinjiang and Inner Mongolia, the sunlight of Qinghai and Gansu, the water of Sichuan and Yunnan... these regions hold abundant, low-cost clean energy. Building intelligent computing centers at the very source of this "green electricity" is like siting a steel mill beside a coal mine and an iron ore deposit. This not only greatly lowers electricity costs but, more importantly, allows the locally generated "curtailed wind," "curtailed solar," and "curtailed hydro" -- energy often wasted because it lies far from load centers -- to be consumed on the spot, transforming unstable green energy into steadily delivered, high-value-added computing services. This is in itself a form of national-level energy optimization.
  2. Climate (solving the "heat source" problem): The western region generally sits at higher altitudes with a cool, dry climate. Guizhou, for example, has an average annual temperature of around 15 degrees Celsius, and Ulanqab in Inner Mongolia is known as the "Prairie Cloud City" with cool summers. These climatic conditions provide a natural "cold source" for intelligent computing centers. By employing natural air cooling, indirect evaporative cooling, and similar technologies, the energy consumed for cooling can be greatly reduced. In traditional air-cooled data centers, cooling systems typically account for 20 to 40 percent of total energy consumption (depending on climate and technology route, with industry surveys commonly citing figures around the upper 30s); in the western region, drawing on natural conditions, this proportion can be dramatically compressed. It is like enjoying nature's air conditioning for free.
  3. Geography and resources: The western region is vast and sparsely populated, with low land costs and relatively stable geological structures, providing ample space for building large-scale data center clusters. Moreover, although water is scarce in some areas (such as the northwest), it is relatively abundant in the southwest, offering different choice paths when we later discuss water usage efficiency (WUE).

The essence of "Eastern Data, Western Computing" is to use high-speed optical fibers in place of ultra-high-voltage transmission lines, effecting a higher-efficiency "West-to-East Electricity Transmission."

In the past, we converted western energy into electricity and sent it eastward over ultra-high-voltage and other long-distance transmission lines, incurring a certain proportion of line loss along the way (typically on the order of single-digit percentages, varying with voltage class and distance). Now, we convert western electricity into computing power locally, and then deliver this higher-order "digital energy" -- computing power -- to the East over fiber optic networks whose transmission loss is far lower than line loss. This is a generational upgrade in the mode of energy transmission.

1.1.3 The Manager's Chessboard: How to Play the "Eastern Data, Western Computing" Move

For a local government manager or the head of a state-owned enterprise, once one understands the thermodynamic nature and the national strategic intent of "Eastern Data, Western Computing," the decision is no longer "should we build in the West," but rather "what should we build in the West, and what should we keep in the East."

This requires us, like chess players, to classify the demand for computing power and place it properly across the two boards of "East" and "West."

"Hot Data" and Eastern Hubs

Not every computing task is suited to the distant West. One category of business, which we call "hot data" processing, is extremely sensitive to network latency.

• High-frequency financial trading: victory or defeat is decided within milliseconds; any additional network latency can mean enormous financial losses.
• Real-time decision-making for autonomous driving: vehicles must communicate with roadside computing units in real time to respond to sudden traffic conditions; the industry generally requires end-to-end vehicle-road latency to stay within the order of 10 milliseconds.
• Remote surgery and VR/AR interaction: the user's real-time experience depends directly on latency; excessive latency brings dizziness and operational errors.
• Emergency response for the city brain: real-time adjustments to traffic signals, for example, must be completed locally in an instant.

For these "hot data" tasks, we should build "edge data centers" or "regional intelligent computing hubs" within or near the eastern city clusters. Their scale need not be the largest, but they must pursue extreme low latency and high reliability. They are the "outposts" and "bridgeheads" on the "Eastern Data, Western Computing" chessboard, responsible for the computing tasks that "cannot wait."

"Cold/Warm Data" and Western Bases

Opposite to "hot data" stand a large number of "cold/warm data" processing tasks. These demand high throughput but are relatively tolerant of real-time requirements.

• Large-scale AI model training: training a model with hundreds of billions of parameters often requires continuous operation for weeks or even months. A few tens of milliseconds of network latency is almost negligible relative to the entire training cycle.
• Film and animation rendering: the post-production rendering of a movie is likewise a massive computing task that can be scheduled in advance and completed in batches.
• Scientific computing and climate simulation: gene sequencing, universe simulation, long-term climate prediction -- these are all typical background, high-throughput computing tasks.
• Data backup and archiving: a quintessentially "cold data" application, entirely insensitive to latency.

These "cold/warm data" tasks are precisely the core workloads that should be migrated to the western intelligent computing bases. Deploying these "computing-intensive" rather than "latency-sensitive" tasks in the West maximizes the benefits of low-cost energy and natural cooling there, achieving a win-win of economic and environmental returns.

A mature national intelligent computing center layout must therefore be an organic system of "East-West linkage, cold-hot separation." The East is like nerve endings -- keenly sensing, swiftly responding; the West is like the cerebral cortex -- deeply thinking, centrally processing. What managers must do is establish an effective scheduling mechanism for computing power that, like an intelligent traffic command system, precisely dispatches each type of computing task to the most suitable node.

If we play this game well, we can truly and continuously turn the West's wind and sunlight into the East's GDP and innovative vigor, and turn the East's vast data into the "new oil" driving the West's green development. This is not merely a financial calculation; it is a decisive opening move toward national energy security, coordinated regional development, and the sustainability of the digital economy -- a move that secures the future.

1.2 The PUE Trap: From a Single Metric to Systemic "Green Accounting" Thinking

In any discussion of building and operating intelligent computing centers, one term appears with such frequency that it has become almost the sole totem for judging whether a center is "green." That term is PUE.

PUE, short for Power Usage Effectiveness, has a very simple formula:

PUE = Total Data Center Energy Consumption / IT Equipment Energy Consumption

The ideal value of this metric is 1.0, meaning that every joule of electricity entering the data center is used 100 percent by IT equipment such as servers, storage, and networking, with nothing wasted on auxiliary systems like cooling and power distribution. Of course, this is impossible in practice. Today, the world's most advanced intelligent computing centers typically achieve PUE values between 1.1 and 1.3. A center with a PUE of 1.2 means that for every kilowatt-hour consumed by IT equipment, an additional 0.2 kilowatt-hours are consumed for cooling and power supply.

The advent of PUE was undoubtedly an important milestone on the data center industry's path toward refined, green management. It provides a simple, intuitive yardstick that allows managers to quantify and compare the energy efficiency of different data centers and to direct technological improvement and optimization accordingly. Over the past decade, the relentless pursuit of an ever-lower PUE has spurred the innovation and adoption of a series of energy-saving technologies, including liquid cooling, indirect evaporative cooling, and high-voltage direct current power supply.

Yet when a single metric is elevated to a pedestal and becomes the sole, overriding standard of assessment, a dangerous "trap" follows.

Blind worship of, and mechanical adherence to, PUE is driving us into a predicament of "metric alienation." We may win a handsome PUE number while losing the larger, more fundamental "green accounting" battle. As stewards of national intelligent computing centers, we must cultivate a systemic mindset that transcends PUE and learn to keep a full "green ledger" spanning water, electricity, and carbon.

1.2.1 The First PUE Trap: Ignoring the "Origin" of the Electricity

The greatest blind spot of the PUE formula is that it concerns itself only with how the electricity entering the data center is distributed, and never asks where that electricity comes from.

In the PUE calculator, one kilowatt-hour of "black electricity" from a coal-fired power plant is held perfectly equivalent to one kilowatt-hour of "green electricity" from a wind turbine or a solar panel.

This yields an absurd situation:

An intelligent computing center built in a large eastern city, with a PUE as low as 1.15 but drawing 100 percent of its power from the municipal grid (composed chiefly of coal-fired electricity), could appear "greener" and "more advanced" on paper than a center built on the western Gobi Desert, with a PUE of 1.25 but obtaining 80 percent of its power directly from an adjacent wind or solar farm.

This plainly defies common sense. The former, though slightly more efficient internally, adds real carbon dioxide to the atmosphere with every kilowatt-hour it consumes; the latter, though somewhat less efficient in its cooling and power-supply systems, rests on a clean energy "base color" and leaves a far smaller carbon footprint on the environment.

The first mindset upgrade for managers is to shift from "PUE supremacy" to "green electricity priority."

From the earliest stages of site selection and planning, we should treat "renewable energy accessibility" as a core consideration on a par with land and network. We should give priority to locations near large wind, solar, and hydro bases, and even explore new "source-grid-load-storage" integrated models, packaging the intelligent computing center together with new energy power stations to achieve direct, "front-of-store, back-of-factory" green power supply.

The benefits of this approach are manifold:

• Environmental benefit: it curbs the carbon emissions of the intelligent computing center at the source, genuinely serving the "dual carbon" goals.
• Economic benefit: by signing long-term power purchase agreements with new energy companies, low electricity prices can be locked in for the next 10 or even 20 years, hedging against the volatility of fossil fuel prices.
• Social benefit: it helps the western region absorb unstable new energy, converting "wasted electricity" into high-value computing services and supporting the transformation of the nation's energy structure.

Accordingly, our assessment system must incorporate a new metric, which we call the "green electricity penetration rate" -- the proportion of a center's total annual electricity consumption that comes from green sources. A truly green intelligent computing center is the combination of a high green-electricity penetration rate and a low PUE. A center with green electricity above 80 percent but a PUE of 1.3 contributes far more to the national "dual carbon" strategy than one with zero green electricity and a PUE of 1.1.

1.2.2 The Second PUE Trap: Hiding the Cost of Water

In the pursuit of an extreme low PUE, many data centers adopt evaporative cooling technology. Its principle is simple: it exploits the physical property that water absorbs a large quantity of heat as it evaporates, using that absorption to cool servers. This technology, particularly in arid regions, achieves far higher cooling efficiency than traditional air-cooled air conditioning and can very effectively bring down the PUE figure.

Yet in the PUE formula, both the numerator and the denominator are "electricity"; it makes no allowance whatsoever for the consumption of "water."

Thus the second trap appears: we may sacrifice vast quantities of precious water for the sake of a handsome PUE number.

In water-rich regions, this may not be a serious problem. But in many areas west of China's "Hu Huanyong Line," water is a scarcer and more precious strategic resource than electricity or land. In such places, blindly launching cooling schemes that consume water is nothing short of "drinking poison to quench one's thirst." A large data center using evaporative cooling can consume water on the order of millions of cubic meters a year (an order-of-magnitude illustration that varies widely with scale and cooling scheme) -- the equivalent of the annual consumption of tens of thousands of households. In a region already arid and water-stressed, this can only aggravate local ecological pressure and sharpen the competition for water between industry and agriculture.

To escape this trap, we must introduce a second key metric: WUE (Water Usage Effectiveness).

WUE has several definitions. One commonly used formula is:

WUE = Annual Total Water Consumption of Data Center (liters) / Annual Total Energy Consumption of IT Equipment (kilowatt-hours)

The unit of this metric is "liters per kilowatt-hour"; it measures how many liters of water must be evaporated or consumed for every kilowatt-hour used by IT equipment. Clearly, the lower the WUE value, the better. A data center with a WUE of zero is a "zero-water" data center.

The second mindset upgrade for managers is to establish a balanced view of "water-electricity dual control."

When selecting technologies, we must not only ask, "What PUE can this solution achieve?" but also press further: "What is its WUE?" We must pay as much attention to local water prices, water resource quotas, and the environmental impact assessment of water extraction as we do to electricity prices.

This requires us to adopt differentiated technical routes tailored to local conditions:

• In water-abundant southwestern regions (such as Guizhou and Sichuan): highly efficient evaporative cooling can be moderately adopted, trading water for electricity in pursuit of an extreme PUE. At the same time, the ecological carrying capacity of the water source must be weighed, to avoid damaging the local water environment.
• In the extremely arid northwestern regions (such as Xinjiang, Gansu, and Inner Mongolia): evaporative cooling schemes must be strictly limited or prohibited. Even if this means a somewhat higher PUE, water resource security must take priority. In these areas, closed cooling systems (such as those using chilled water and cooling towers, where water circulates within the system with only minimal evaporative loss) and advanced liquid cooling should be vigorously promoted. Liquid cooling in particular, by bringing the coolant into direct contact with the chip, dissipates heat far more efficiently than air, greatly reducing the reliance on fans and air conditioning and thereby achieving a very low PUE without consuming water.
• Exploring "water-heat" combined use: a still more systemic line of thinking is to consider how the "wastewater" (treated cooling water) and "waste heat" discharged by the intelligent computing center can be turned to resourceful use. Warm cooling water, for example, could irrigate and heat the surrounding agricultural greenhouses, or be fed into the municipal heating network to supply winter heating to residential areas. This "computing-agriculture" and "computing-livelihood" coupling transforms the intelligent computing center from a mere energy consumer into an active participant in the regional energy cycle.

Introducing the WUE metric is, in essence, a demand that we expand the environmental impact assessment of the intelligent computing center from a single "energy" dimension to a two-dimensional plane of "energy-water resources." A truly environmentally responsible intelligent computing center must find an optimal balance point on this two-dimensional plane, rather than pursuing the extreme along any single dimension.

1.2.3 The Third PUE Trap: Concealing the Full Picture of Carbon

Even if we consider both PUE (the efficiency of electricity) and green-electricity penetration (the source of electricity), and supplement them with WUE (the consumption of water), we remain one last -- and most critical -- step short of a complete "green ledger."

That step is answering an ultimate question: from cradle to grave, how many greenhouse gases does the construction and operation of this intelligent computing center emit into the earth altogether?

This is the concept of CUE (Carbon Usage Effectiveness).

The formula for CUE is:

CUE = Total Carbon Emissions of Data Center (kg CO2 equivalent) / Total Energy Consumption of IT Equipment (kilowatt-hours)

The unit of this metric is "kg CO2 equivalent per kilowatt-hour"; it measures the total carbon emissions corresponding to every kilowatt-hour consumed by IT equipment. Here too, the lower the CUE value, the better.

The third trap of PUE is that it is a "process indicator" rather than a "result indicator." It measures only the energy efficiency of the operational phase, while completely overlooking the enormous "embedded carbon" generated in the other stages.

A complete carbon footprint of an intelligent computing center (the numerator in the CUE formula) should encompass at least three parts, which in carbon accounting systems are known as "Scope 1, Scope 2, and Scope 3":

Scope 1: Direct Emissions

This part refers chiefly to emissions from the direct combustion of fossil fuels within the data center, the most common source being backup diesel generators. Although rarely used, their testing and emergency operation across the full lifecycle also produce carbon emissions.

Scope 2: Indirect Emissions (Electricity)

This is the largest share of carbon emissions -- the emissions arising from the consumption of purchased electricity, as discussed earlier. Its calculation is directly tied to the green-electricity penetration rate. The more "black electricity" consumed, the higher these emissions. A center with a very low PUE but running entirely on thermal power will have staggering Scope 2 emissions.

Scope 3: Indirect Emissions (Supply Chain and Construction)

This is the part most easily overlooked, most difficult to account for, and yet potentially enormous in volume. It includes:

• Upstream supply chain emissions: the carbon emitted in their own factories during the production of all materials -- servers, chips, switches, optical modules, cement, steel, and so on. The very fabrication of a high-performance GPU is itself energy-intensive and carbon-intensive.
• Construction-phase emissions: the carbon emitted during civil engineering, equipment transportation, installation, and commissioning.
• Downstream waste-processing emissions: the carbon released in the recycling, dismantling, and disposal of electronic equipment such as servers once they reach the end of their service life.

The third mindset upgrade for managers is to establish a "full-lifecycle" carbon accounting awareness.

We cannot smugly gaze at the handsome PUE figure on the operations report while shutting our eyes to the mountains of steel and cement consumed during construction, or to the "manufacturing carbon" lurking behind the thousands of servers in the halls.

Establishing this systemic mindset will fundamentally reshape our decision-making:

1. In equipment procurement: we must look not only at a server's performance and price but also demand from suppliers their products' "carbon footprint reports." Where performance is comparable, we should give priority to "green servers" that emit less carbon in manufacturing and incorporate more recycled materials. Through such procurement decisions, we exert green pressure upstream, driving the low-carbon transformation of the entire IT industry chain.
2. In architectural design: promote green building materials and prefabricated, modular construction to reduce waste and emissions at the construction site; design for a longer building service life and avoid frequent demolition and rebuilding.
3. In asset management: extend the service life of servers. In the past, the industry was accustomed to retiring a batch of servers every three to five years. But if, through software optimization and maintenance, we can stretch their productive life to six or seven years, we effectively amortize their enormous "manufacturing carbon" -- a great contribution to the environment. At the same time, we should establish a comprehensive system for recycling and reusing e-waste.
4. In final assessment: make CUE the "ultimate indicator" of a center's green level. PUE, WUE, and green-electricity penetration are all optimization parameters along the way; ultimately they should all serve the overarching goal of reducing CUE. A local government that takes the lead in establishing a CUE-based green assessment standard for intelligent computing centers will have seized the discourse power in standard-setting for the nation's green digital economy.

Conclusion: From "Small Accounting" to "Great Accounting"

This chapter, setting out from the physical essence of "computing power is heat," has step by step laid bare the limitations of the traditional models of site selection and energy evaluation for managers.

• Small economic accounting: looking only at land prices and electricity prices. This is a short-sighted, opportunistic mindset that will trap us in the "around the capital, around Shanghai" layout and ultimately lead to unsustainability.
• Great national accounting: grasping the strategic connotation of "Eastern Data, Western Computing" and regarding it as a national-level optimization of energy and computing resources. This demands a holistic view, an understanding of "cold-hot separation," and a rational layout.
• The "efficiency accounting" of PUE: this is progress in refined management, but it easily falls into the trap of "metric alienation."
• The "great green accounting" of PUE + green-electricity penetration + WUE + CUE: this is a systemic, full-lifecycle mindset of sustainable development. It requires us to bring electricity, water, carbon, supply chain, and construction -- all the elements -- into a unified framework for comprehensive consideration.

As infrastructure stewards of the new era, our responsibility is by no means merely to erect data centers that score well on benchmarks and turn a profit. Our historical mission is to forge for the nation a digital foundation that is truly "environmentally friendly, resource-conserving, autonomous, and controllable."

This foundation must be deeply rooted in the geography and resource endowments of our country. It must be able to transform the wind and sunlight of the West into the wisdom of the East, and, while delivering surging computing power, to keep our lucid waters and lush mountains safe.

The first step in sharpening the "eye of observation" is to cast a more far-reaching, more systematic, more responsible gaze upon the land beneath our feet, and to find for our "convergence engine" the most solid, the greenest, and the wisest place to stand. This is not merely accounting. It is laying out the future and laying the foundation for civilization.