The debate has moved from software to infrastructure

On August 24, Reuters reported that political opposition to artificial intelligence data centers had begun to affect investor sentiment toward technology companies. The shift shows that AI expansion is no longer assessed only through model capability: power availability, grid infrastructure, and relationships with local communities have entered the decision.

In Texas, the state government ordered an audit on August 3 of projects waiting to connect to the grid. The official statement says the queue considered by ERCOT exceeded 474 gigawatts, more than five times the system record demand. That amount represents connection requests under review, not contracted consumption, but it reveals the scale of the task of separating viable projects from proposals that have not yet been demonstrated.

Power, water, and local costs are now part of the product

The standards announced by Texas require new data centers to demonstrate that they can meet connection conditions, protect grid reliability, and avoid transferring required infrastructure costs to residents. The state also included water conservation, noise, and neighborhood impact among the assessment criteria.

ERCOT preliminary forecast places total demand near 367. 8 gigawatts in 2032, compared with the historical peak of 85. 5 gigawatts recorded in 2023. The forecast covers all economic growth in the state, not data centers alone, but it helps explain why large loads must be verified before entering system planning.

At a national level, the International Energy Agency estimated that data centers consumed about 180 terawatt-hours in the United States in 2024 and that consumption may grow by roughly 240 terawatt-hours by 2030. Digital expansion therefore depends on physical decisions made long before a person opens an AI tool.

What changes for companies that use artificial intelligence

For most companies, the conclusion is not to build a power plant or stop AI projects. It is to include infrastructure, cost, and continuity in the value assessment. Larger models are not always necessary for simple tasks, and poorly designed processes can consume more capacity without producing a better result.

Supplier selection also brings new questions: where the service operates, how it handles demand peaks, what commitments it makes about power and water, and how it maintains availability when the grid is constrained. These factors are no longer only environmental topics; they can affect price, schedule, reputation, and operational continuity.

How to grow without turning scale into waste

A responsible strategy connects every AI workload to a verifiable outcome and considers efficiency from the design stage. The goal is not to restrict innovation, but to ensure scarce capacity is used where it creates real value.

• Use a model suited to the task rather than always choosing the largest option.
• Measure cost, time, consumption, and quality for each automated process.
• Plan peaks, contingency, and continuity before expanding the workload.
• Require supplier transparency on location, power, water, and availability.
• Review data and workflows to avoid reprocessing, duplication, and calls that produce no useful result.

What changed

The Dutch data protection authority imposed a €824. 99 million fine on Uber. According to the regulator, between 2018 and 2022 the platform used systems that suspended driver accounts, sometimes permanently, without adequate information and without a person checking for possible mistakes.

The decision concerns automated processing: software analyzes data and makes a decision without meaningful human participation. Uber disputes both the findings and the size of the fine, says the reviewed practices are historic, and states that its current processes include human review and an opportunity to appeal; the company plans to challenge the decision.

The lesson is not that every automated process is improper. It is that machine speed also multiplies the effect of poor data, incomplete rules, or signals interpreted without context when the outcome can interrupt someone’s income.

Why this matters

For a platform driver, losing account access can mean instantly losing the tool used to earn a living. In other industries, similar decisions can stop a payment, reject a credit application, remove a candidate from hiring, or deny access to an essential service.

In Europe, the rule invoked in the case protects people from fully automated decisions with legal or similarly significant effects, except in specific circumstances. In Brazil, the data protection authority notes that the LGPD gives people a right to request review and explanations of the criteria and procedures used; in practice, human review is meaningful only when the reviewer understands the case, can see relevant information, and has real authority to change the outcome.

Consider a fraud system that freezes an account after detecting an unusual location change. Without context, it may treat legitimate travel as a risk; without an explanation and a way to challenge the result, the person cannot correct the data or show what happened. Operational efficiency then becomes a human and reputational cost.

Reliable data comes before automation

Every decision system depends on what it receives: ratings, history, location, documents, and risk signals. Data cleansing means finding duplicates, stale records, inconsistent fields, and missing context before those records feed rules or models. It is not simply tidying a database; it reduces the chance that an operational error becomes real harm.

Organizations also need to know where each data point came from and record how it contributed to a decision. That trail helps teams investigate false positives — legitimate cases incorrectly flagged as problems — compare outcomes across groups, and explain the result in language a person can understand.

The greater the consequence, the higher the quality threshold should be. A content-ranking system may tolerate lightweight corrections; a system that changes income, credit, health, or access to rights needs testing, monitoring, documentation, and fast paths to reversal.

What companies can learn

Governance should not arrive only after the automation is finished. A Technology Cell can bring product, data, operations, security, legal, and support together to decide what may be automated, what requires human confirmation, and how affected people will be informed.

The goal is not to slow every innovation, but to decide where speed is safe. When an organization combines cleansed data, clear limits, and genuine human review, automation gains something a model cannot provide alone: legitimacy in situations that affect people.

• Map automated decisions and classify the impact of each one.
• Validate data quality, origin, and context before automating.
• Explain relevant factors and provide an accessible challenge process.
• Give human reviewers the information, time, and authority to correct outcomes.
• Monitor errors, reversals, and unequal effects over time.

The spectacle shows how quickly performance is advancing

The World Humanoid Robot Games opened in Beijing with 2,056 robots from 666 teams across 51 events, according to the city government. The program combines races and sports with tasks inspired by factories, restaurants, offices, and emergency situations.

Reuters reported on August 23 that two robots completed the 100 metres faster than the 9. 58-second human world record. Another model covered 400 metres in 39. 7 seconds. The figures attract attention and show rapid progress in movement, energy management, and control.

The real challenge appears in small imperfections

Speed alone does not solve problems in the physical world. After the finish line, some robots still needed a padded barrier to stop. In everyday work, a cable at the wrong angle, a shifted box, or an object slightly out of reach can require perception, coordination, and error recovery.

More than 40% of the events require fully autonomous operation, according to information cited by Reuters. Tests include connecting cables, loading materials, handling warehouse operations, serving in restaurants, and responding to incidents. These less spectacular situations are where practical usefulness begins to be demonstrated.

Why this matters to organizations

For an organization, the lesson extends beyond humanoid robots. A controlled demonstration can reveal potential, but daily operations include incomplete data, changing environments, exceptions, and people with different needs.

Before expanding physical or digital automation, a company needs to define the expected outcome, decision boundaries, minimum data quality, and a safe procedure for unexpected situations. Technology creates value when it can work consistently inside the real process.

How to separate a demonstration from practical value

A useful evaluation does not ask only whether the technology can perform a task once. It observes repetition, safety, cost, recovery, and the impact on the people who use the process.

• Test the solution in the environment where it will actually be used.
• Measure success rate, time, errors, and the need for human intervention.
• Include variation and exceptions from the beginning of the pilot.
• Define how to stop, review, and safely resume the operation.
• Compare the result with the total cost of implementation and support.

What changed now

NASA announced on August 21, 2026, that the Nancy Grace Roman Space Telescope had completed its flight readiness review. During this formal checkpoint, managers from NASA, the mission, and SpaceX assessed the observatory’s status and certified it to begin final launch preparation activities.

Launch is targeted for no earlier than August 30 at 7: 26 a. m. Eastern time aboard a Falcon Heavy rocket. Before then, the observatory, already protected inside the rocket fairing, will be moved to the launch pad hangar, attached to the vehicle, and transported to Launch Complex 39A.

The review does not guarantee liftoff on that exact date because technical and weather conditions can still change the schedule. The new development is that the project has passed the formal control needed to move from preparation into launch operations.

Why this matters

Roman was built to observe large areas of the sky with detail comparable to Hubble. Its infrared field of view will be about 200 times larger, shifting the work from photographing small patches toward creating broad, repeated maps of the universe.

Those maps will help researchers study how the universe expanded, where dark matter is concentrated, and what kinds of planets exist beyond our solar system. Roman will also carry a coronagraph, an instrument that suppresses the light of a star so much fainter nearby objects, including planets and planet-forming disks, can be observed.

The practical result is faster discovery. Rare events such as exploding stars, gravitational lenses, and planets revealed through tiny changes in light become easier to find when an observatory tracks millions or billions of objects in a consistent way.

The challenge does not end at the camera

Official mission material estimates that Roman will transmit about 1. 4 terabytes of science data per day, more than 500 terabytes per year, and as much as 20 petabytes during its five-year primary mission. One petabyte equals one thousand terabytes. At that scale, downloading the entire collection to a personal computer is no longer practical.

That is why the scientific infrastructure was prepared before launch. The Roman Research Nexus brings data, computing capacity, and analysis tools together in a cloud environment. Instead of moving enormous collections of files, researchers bring their methods to the place where the data is stored.

Official mission data will also be released without an exclusive period for a single team. A group in Brazil, for example, will be able to query selected parts of the archive, combine observations, and collaborate internationally without keeping a local copy of the whole mission. Open access expands participation, but only if files, metadata, versions, and quality rules remain consistent.

What businesses can learn

Roman offers a lesson well beyond astronomy: when volume grows, cleaning data only after it arrives becomes expensive and slows down its use. Quality, standardization, traceability, access rules, and context need to be designed alongside the product or service that creates the information.

This is the natural connection to Data Sanitation. Removing duplicates and reconciling incompatible formats matter, but the larger goal is to create a foundation that people and systems can trust. Without it, artificial intelligence and automation merely process inconsistencies at greater speed.

There is an organizational lesson as well. Missions of this scale connect specialists in instruments, operations, data, and research around a continuous delivery. In a company, a Technology Cell can play a similar role by combining different skills, turning a priority into an operating capability, and improving the system as real-world use produces new knowledge.

What is changing in technology contracts

A Reuters report published on August 20 describes Indian technology service providers moving from contracts based on hours and team size toward agreements tied to performance. The shift is taking place in an industry estimated at $315 billion as clients press for greater productivity and lower costs.

This does not mean every project will adopt the same commercial model. It shows that artificial intelligence is pushing clients and suppliers to define the expected result, how it will be measured, and who carries the risk when the promise is not achieved.

Outcomes must be defined before the technology

Outcome-based pricing sounds straightforward, but it requires a reliable baseline. Faster service, less rework, or greater availability can only be demonstrated when the company understands current performance and agrees on how progress will be measured.

Without consistent data and acceptance criteria, a business may replace one imperfect metric, such as hours worked, with another fragile measure. The contract should record scope, exceptions, expected quality, and human accountability in addition to the main indicator.

Human work moves to a different position

The trend does not remove the importance of people. It shifts more value toward understanding the problem, reviewing decisions, organizing business knowledge, and validating what automation produced. Smaller teams may gain speed, but experience remains essential when real situations move beyond the pilot.

In an analysis published on August 12, OpenAI reports that companies are moving from AI as assistance toward workflows in which agents execute parts of the work. The analysis also recommends appropriate context, clear permissions, governance, and human review to turn individual uses into repeatable processes.

How to experiment without overpromising

A safer approach is to choose a bounded process, measure the starting point, and run a pilot with clear accountability. Only after observing quality, cost, adoption, and unexpected effects should the organization expand automation or connect payment to the result.

• Define a business outcome that can be measured without ambiguity.
• Record the baseline, data sources, and accountable owners.
• Set acceptance criteria, human review, and exception handling.
• Track errors, rework, total cost, and impact on users.
• Review the contract when the context or data changes.

What Brazil announced

Brazil’s National Laboratory for Scientific Computing said it is leading the deployment of a new artificial intelligence supercomputer at the Augusto Severo Science and Technology Park in Macaíba, Rio Grande do Norte. The public selection estimates about R$ 959 million for the integrated solution within a broader set of federal AI infrastructure initiatives.

Why a supercomputer matters

Advanced AI models require substantial capacity to learn from large volumes of information and then respond to new requests. National infrastructure could support universities, public agencies, and innovation projects that currently depend on scarce computing resources or capacity contracted abroad.

Data is infrastructure too