RealityBreaks

Visuals for Business, Socials, Education & Fun
Learn to use Ai

AI News Roundup: AI Is Becoming Cheaper to Use—and Far More Expensive to Build

Written by

Elliot Gray

Hello, I’m Elliot Gray from RealityBreaks.

This week’s artificial-intelligence news reveals a fascinating contradiction.

For ordinary users and smaller businesses, powerful AI is becoming faster, cheaper and easier to access.

Behind the scenes, however, the companies and governments building that technology are committing tens—or even hundreds—of billions to computer chips, data centres, electricity networks and cloud infrastructure.

We also saw another leap forward in AI-generated video and a reminder that realistic synthetic media is beginning to create difficult questions for elections, advertising and public trust.

Here are the most relevant stories reported during the week ending Friday, July 31, 2026, explained in plain English.

1. China’s MiniMax Launches H3, a New AI Video Model With Sound

Chinese AI company MiniMax launched H3 on July 31, a new multimodal video-generation model that can work with text, images, existing video and audio.

The company says H3 can generate videos lasting up to 15 seconds in 2K resolution with native stereo sound. It also supports video editing and motion transfer, allowing movement from one clip to be applied to another subject or scene. MiniMax plans to release the model’s weights, giving developers greater scope to customise it. Reuters

Beginner-Friendly Explanation

Most earlier AI video generators concentrated on creating moving pictures from a written prompt.

Newer systems are becoming more like complete creative-production tools.

A user may be able to provide:

  • a written description;
  • a reference photograph;
  • an existing video;
  • a piece of audio;
  • or a combination of these.

The model then uses those materials to create or edit a finished clip.

“Native sound” means the audio is generated as part of the video rather than being added separately afterwards. That can include ambience, effects, dialogue or music, depending on the model’s capabilities.

Why This Matters

AI video is rapidly moving from short experimental clips toward practical commercial production.

MiniMax says H3 is aimed at industries including:

  • advertising;
  • ecommerce;
  • product design;
  • gaming;
  • social-media content.

The company also claims the model’s generation cost is less than one-third of that charged by mainstream competitors. That claim will need independent testing, but lower prices would make professional-looking video more accessible to smaller organisations. Reuters

The decision to make model weights available is also significant. Most leading video systems remain tightly controlled by their developers. A more open model could allow businesses and creators to build specialised tools around it.

Practical Takeaways

For Individuals

AI video quality is improving quickly, but do not judge a model only by promotional demonstration clips.

Test whether it can consistently produce:

  • believable movement;
  • accurate faces;
  • readable text;
  • reliable character continuity;
  • suitable sound;
  • commercially usable resolution.

Always check the provider’s terms before using generated content commercially.

For Small and Medium-Sized Businesses

AI video may now be worth testing for:

  • product demonstrations;
  • social-media adverts;
  • short explainers;
  • service introductions;
  • property or travel visualisation;
  • training materials.

Begin with one defined use case and compare the cost against conventional production.

Keep a human involved in checking brand accuracy, factual claims, product details and any representation of real people.

2. Europe Commits €10 Billion to Seven AI Gigafactories

The European Union announced plans on July 30 to fund seven AI gigafactories across the bloc with €10 billion of public funding.

The European Commission hopes to attract at least another €20 billion in private investment. Each proposed facility is expected to combine advanced AI processors, cloud systems, software, high-speed connectivity and large-scale data-centre capacity. Successful projects are expected to be selected in early 2027. Reuters

Beginner-Friendly Explanation

An AI gigafactory is not a factory producing robots on an assembly line.

It is an enormous computing facility designed to train and operate advanced AI models.

The proposed European facilities are expected to contain at least 100,000 high-performance AI chips each. That computing power would be used by researchers, startups, public organisations and larger companies developing AI services. AP News

Why This Matters

Europe has strong universities, engineering talent and industrial companies, but much of the world’s leading AI computing infrastructure is controlled by American technology firms.

European policymakers are concerned that relying too heavily on overseas cloud platforms could create problems involving:

  • economic competitiveness;
  • data security;
  • access to computing power;
  • pricing;
  • technological independence.

The Giga factory project is an attempt to give European companies more access to the infrastructure required to develop advanced AI.

However, the facilities will also need enormous quantities of electricity and cooling. Europe’s comparatively high energy prices could make the programme more difficult and expensive than equivalent expansion in other regions. AP News

Practical Takeaways

For Individuals

This investment will not immediately produce a new chatbot for you to use.

Its effect will be longer-term. It could help European researchers and companies develop more home-grown AI tools and reduce reliance on a small group of overseas providers.

For Small and Medium-Sized Businesses

European SMEs may eventually gain better access to:

  • regional cloud infrastructure;
  • specialised industry models;
  • EU-compliant AI services;
  • research partnerships;
  • public funding and innovation programmes.

Businesses should not wait for these factories before using AI, but they should monitor future European support schemes.

3. OpenAI Cuts the Cost of GPT‑5.6 and Pushes Faster Processing

OpenAI announced major price reductions for two members of its GPT‑5.6 model family on July 30.

The company reduced the price of GPT‑5.6 Luna by 80% and GPT‑5.6 Terra by 20%. It also introduced a faster API processing option for GPT‑5.6 Sol that can deliver responses up to 2.5 times faster than standard processing, although at twice the standard price. OpenAI

OpenAI says the smaller Luna and Terra models are designed to provide stronger performance per pound or dollar for routine business workloads, including workplace questions, coding support and automated agent tasks. OpenAI

Beginner-Friendly Explanation

The largest and most intelligent AI model is not always the best model for every job.

Think of transport.

You might need a heavy lorry to move industrial equipment, but using that lorry to deliver a single envelope would be wasteful.

Smaller AI models can often handle routine work more quickly and cheaply, including:

  • sorting information;
  • classifying emails;
  • extracting details from documents;
  • producing short summaries;
  • answering standard questions;
  • carrying out repetitive background tasks.

Why This Matters

Cost has been one of the biggest barriers to using AI at scale.

A person occasionally using a chatbot may not notice the expense. A business processing thousands of customer enquiries or documents certainly will.

Lower-cost capable models could make it practical for smaller companies to automate processes that were previously too expensive.

It also shows how fierce the model market has become. AI providers are now competing not only on intelligence, but on speed and cost per completed task.

Practical Takeaways

For Individuals

Do not assume that selecting the most advanced model always produces better value.

For everyday writing, planning, summarisation and organisation, a smaller model may be sufficient.

Reserve the most powerful options for tasks involving complex reasoning or extensive research.

For Small and Medium-Sized Businesses

Divide AI work into levels.

Routine tasks may use a smaller, cheaper model.

Important analytical tasks may require a stronger model.

High-risk tasks involving legal, financial, employment or sensitive customer decisions should also receive qualified human review.

Measure the cost of the completed process rather than merely comparing the advertised price per token.

4. Amazon Raises Planned Technology Spending to $220 Billion

Amazon announced on July 30 that it expects to spend around $220 billion during 2026, increasing its previous capital-investment plan by $20 billion.

Much of this spending is directed toward AI infrastructure, cloud computing, custom chips, robotics and related technologies. Amazon Web Services reported 37% year-on-year growth during the second quarter, while Amazon said demand for AI computing remains so strong that it still expects capacity constraints despite the increased spending. AP News

Beginner-Friendly Explanation

When people use an AI service, the work is usually performed in a remote data centre.

Cloud companies such as Amazon rent that computing power to AI developers and other businesses.

Amazon is therefore investing in:

  • more data centres;
  • more servers;
  • specialist AI chips;
  • networking equipment;
  • electricity capacity;
  • cooling systems.

The important phrase is capacity constraint.

It means customer demand for computing power is increasing faster than Amazon can build the infrastructure needed to provide it.

Why This Matters

The extraordinary level of investment suggests that large technology companies do not see AI as a temporary trend.

They are building infrastructure intended to support years of demand.

But the spending also creates risks.

Companies must eventually earn enough revenue from AI and cloud services to justify these enormous commitments. Investors are increasingly questioning whether every company will receive an adequate return from its infrastructure expansion. AP News

For customers, shortages could affect service availability, usage limits and prices.

Practical Takeaways

For Individuals

Apparently effortless AI relies on expensive physical infrastructure.

Free services may eventually introduce tighter limits, more advertising, premium upgrades or new ways of generating revenue.

Review privacy settings and understand how a free service is funded.

For Small and Medium-Sized Businesses

Avoid paying for several overlapping AI platforms without checking whether staff use them.

Review each subscription regularly:

  • What task does it solve?
  • How often is it used?
  • How much time does it save?
  • Does another tool provide the same feature?
  • Could a smaller model do the work more cheaply?
  • Can your data and workflows be exported?

Do not let an important process become completely dependent on one provider.

5. An AI Avatar Creates New Questions for Elections and Public Trust

Former Brazilian president Jair Bolsonaro, who was under house arrest, appeared through an AI-generated avatar at the launch of his son Flavio Bolsonaro’s political campaign on July 25.

A Brazilian law professor quoted by Reuters described the appearance as a clear violation of electoral rules. The incident illustrates how AI-generated representations of real people can be used to create appearances or messages even when the individual is not physically present.

Beginner-Friendly Explanation

An AI avatar is a synthetic version of a person.

It may copy the individual’s:

  • face;
  • voice;
  • expressions;
  • speaking style;
  • body movement.

Some avatars are created with permission for harmless or useful purposes, such as translating a business presentation into another language.

The problem arises when audiences cannot easily tell whether the person genuinely recorded, approved or even said what appears in the video.

Why This Matters

AI-generated media is becoming realistic enough to influence public opinion.

This creates risks involving:

  • political misinformation;
  • impersonation;
  • fraudulent endorsements;
  • reputational damage;
  • fabricated evidence;
  • misleading advertising.

The issue is not limited to national politics.

A fake video of a company director, local business owner or employee could be used to request a payment, damage a reputation or promote a fraudulent service.

Practical Takeaways

For Individuals

Do not assume that a realistic video proves a person genuinely said or did something.

Before sharing unusual footage:

  • check the original source;
  • look for reporting from reliable outlets;
  • inspect whether the account publishing it is genuine;
  • search for an unedited version;
  • be cautious when the content provokes immediate anger or urgency.

For Small and Medium-Sized Businesses

Create a basic impersonation-response plan.

Staff should know how to verify unexpected video, voice or email instructions involving:

  • payments;
  • bank details;
  • passwords;
  • confidential documents;
  • changes to suppliers;
  • urgent executive requests.

Use a second communication method before acting on unusual financial instructions.

Businesses using avatars themselves should state clearly that the content is AI-generated and confirm that the represented person has approved its use.

Elliot Gray’s Closing Perspective

This week’s stories show the AI industry developing at two very different speeds.

At the user level, the technology is becoming cheaper and easier to access.

OpenAI is reducing model prices. MiniMax is promising lower-cost video production. AI features are becoming available to businesses that could never have afforded advanced creative or analytical systems a few years ago.

At the infrastructure level, however, the numbers are becoming almost difficult to comprehend.

Europe is committing billions to computing facilities. Amazon is preparing to spend $220 billion in a single year. Technology companies are competing for chips, electricity, land and data-centre capacity.

And as AI-generated content becomes more convincing, society is being forced to answer a separate question:

Can we still trust what we see and hear?

For individuals, the best response is not fear.

It is awareness.

Learn what these tools can do, check important information and avoid giving unfamiliar services unnecessary access to your data.

For small and medium-sized businesses, the opportunity is substantial—but good judgement matters more than excitement.

Use the lowest-cost tool that reliably performs the task. Measure whether it creates genuine value. Protect confidential information. Keep a human responsible for important decisions.

The strongest AI strategy is not about adopting everything.

It is about choosing carefully, using responsibly and improving gradually.

Call to Action

Want practical artificial-intelligence guidance without the jargon, panic or exaggerated promises?

Visit the RealityBreaks AI Blog for weekly news roundups, beginner-friendly guides, creative AI tutorials and practical advice for using AI safely at home and in business.

For a structured introduction designed particularly for people who feel technology may be moving too quickly, explore AI After 50: The Plain-English Guide to Using Artificial Intelligence.

More posts