DEV Community: Tekk NG

AI Data Centres Are Running Into a Power Problem. What Does That Mean for Tech?

Tekk NG — Mon, 19 Oct 2026 10:55:43 +0000

AI infrastructure discussions often focus on GPUs, model architectures, inference costs, and cloud platforms. But there is another dependency behind every large AI workload: electricity.

Hyperscale data centres need power for compute, networking, storage, and cooling. As AI adoption increases, some regions are finding that their existing electricity infrastructure cannot accommodate new facilities quickly enough.

This is turning power availability—and the question of who pays for grid upgrades—into a major issue for the technology industry.

The infrastructure behind an AI request

Consider what happens when an AI application receives a request.

The application may send data to a backend service, which routes the workload to an inference server running on specialised hardware. That server consumes electricity, while the wider facility must provide cooling, networking, backup power, and monitoring.

At scale, thousands of servers operate together. Even relatively efficient individual components can add up to substantial facility-wide demand.

This is why AI infrastructure planning involves more than choosing a cloud provider or purchasing accelerators. Power capacity and cooling can affect the location, cost, and operational readiness of a deployment.

Why grid connections are becoming a bottleneck

When a proposed data centre requires more power than a local network can provide, utilities may need to upgrade substations, transformers, transmission lines, or other infrastructure.

Those upgrades involve planning, financing, equipment procurement, construction, and regulatory approval. The process can take considerably longer than deploying servers inside an existing building.

The consequences include delayed projects, higher infrastructure costs, and uncertainty about when planned computing capacity will become available.

A recent example emerged in the UK. On October 18, 2026, reports said Microsoft had withdrawn from a major data-centre project in Loughton, Essex, amid delays to the site's electricity connection. Sufficient grid capacity was reportedly not expected until the 2030s.

The lesson for infrastructure teams is straightforward: compute capacity is useful only when the supporting infrastructure is ready.

Who should pay for the upgrades?

The issue has become a regulatory priority in the United States.

In June 2026, the Federal Energy Regulatory Commission (FERC) directed six regional grid operators to justify or reform their rules for connecting large electricity users, including data centres.

One concern is the risk of shifting costs to ordinary electricity customers. If a utility builds capacity for a proposed facility and the project is delayed, scaled back, or abandoned, regulators want arrangements that address who remains responsible for the cost.

There is an important distinction between a dedicated connection and wider network investment. A data-centre operator may pay for equipment built specifically for its facility, while broader grid upgrades can fall under different regulated cost-sharing arrangements.

The applicable rules depend on the region and the project.

Three technical and operational considerations

1. Measure energy efficiency, not just compute throughput

AI infrastructure teams should consider the energy required to deliver useful work, alongside latency, throughput, and cost.

Improving model efficiency, reducing unnecessary inference, scheduling workloads intelligently, and increasing hardware utilisation can all help reduce wasted resources. The actual impact depends on the workload and deployment.

2. Design for realistic capacity constraints

Infrastructure planning should account for power availability, cooling capacity, redundancy, and the time required to obtain a grid connection.

A project plan based only on server delivery dates can miss a critical dependency: whether the site can supply the required power when the hardware arrives.

3. Make infrastructure commitments reliable

Data-centre demand forecasts influence utility investment. Overestimating demand or maintaining speculative connection requests can complicate capacity planning.

More credible forecasts, clear contracts, and appropriate financial commitments can help reduce the risk of building infrastructure that is not ultimately used.

What developers should take away

This is not just an energy-policy issue. It affects cloud pricing, deployment timelines, AI capacity planning, infrastructure architecture, and the long-term economics of running software.

As AI systems grow, engineers will need to think about efficiency and resource constraints across the full stack—not only the application code.

For developers and aspiring technologists looking to build practical skills and understand the broader technology ecosystem, TEKHUB Nigeria is one resource worth exploring.

Final thought

The AI boom depends on more than faster chips and better models. It also depends on power networks that can deliver electricity reliably, at the right place and time, under clear and fair cost arrangements.

What do you think will become the biggest constraint on AI infrastructure: access to electricity, the cost of computing hardware, or the ability to build data centres quickly enough?

AI #CloudComputing #DevOps #Infrastructure #Energy

Apple HomeView: What Developers Should Know About Apple's New AI Smart Home Hub

Tekk NG — Thu, 15 Oct 2026 11:58:57 +0000

Apple has entered the smart display conversation with HomeView, a 7-inch smart home display powered by Siri AI. Announced on October 13, 2026, the device combines smart home controls, personalized user experiences, communication, and AI assistance in one interface.

For developers, the interesting part isn't just the hardware. It's the system behind it: AI-assisted interaction, device integration, user identity, privacy, and the challenge of making connected systems work reliably.

What is Apple HomeView?

HomeView is designed to act as a central interface for a connected home. According to Apple's announcement, it supports:

Voice and touchscreen interaction through Siri AI.
Controls for compatible smart home accessories.
Camera feeds and home monitoring.
FaceTime, messages, calendars, and reminders.
Personalized home screens for up to six recognized users.

The device can sit on a tabletop or be mounted on a wall. Apple has positioned it as a shared household device that can still provide personalized experiences.

1. AI is becoming part of the interface layer

Traditional interfaces expect users to navigate menus, open applications, and select controls. AI-assisted interfaces increasingly let users describe what they want in natural language.

For example, a user might ask an assistant to adjust the lights or check a supported camera feed. The system must interpret the request, determine the intended action, check what it is allowed to do, and communicate with the relevant device or service.

That creates several engineering challenges:

Intent recognition: Does the system understand what the user means?
Context management: Does it have the right information to respond appropriately?
Authorization: Is this user permitted to perform the requested action?
Error handling: What happens when a connected device is offline?
Feedback: Can the user tell whether an action actually succeeded?

A conversational interface is only as useful as the reliability of the systems executing its commands.

2. Device interoperability still matters

A smart home hub needs to communicate with a variety of devices and services. Supporting a broad range of accessories is therefore an important part of the user experience.

For developers working on IoT systems, this highlights the value of well-defined interfaces, documented protocols, robust device discovery, and predictable error responses.

It also reinforces a familiar lesson from software engineering: integration is often harder than building an individual component.

A light switch, camera, or thermostat may work perfectly on its own, but the complete experience can fail if authentication expires, network connectivity drops, or two systems interpret device states differently.

3. Personalization introduces security responsibilities

Apple says HomeView can recognize up to six household members and personalize the experience for each person.

That makes identity and access control especially important. A shared screen may display personal messages, reminders, or other information that should not automatically be available to every visitor.

Developers building similar systems should consider:

Authentication and authorization as separate concerns.
Least-privilege access to devices and personal data.
Clear separation between private and shared household information.
Secure handling of camera feeds and account credentials.
Safe defaults when the system cannot confidently identify a user.
Transparent privacy controls and understandable permission prompts.

Face recognition or voice recognition should not automatically be treated as proof that every sensitive action is authorized. The level of verification should match the risk of the action.

4. Connected devices need dependable failure handling

Smart home software operates in the real world, where networks fail and physical devices behave unpredictably.

A robust system should communicate when a command cannot be completed, avoid claiming success before receiving confirmation, and provide a safe recovery path. For physical actions such as unlocking a door, the system also needs to account for the consequences of an incorrect command.

These principles apply well beyond smart displays. They are relevant to IoT applications, backend services, mobile apps, and distributed systems.

What does HomeView mean for developers?

HomeView is another example of AI moving from standalone software into everyday hardware. Building products in this space requires more than a language model or a polished interface. It requires coordination between hardware, software, networking, identity, and security.

For developers and aspiring technologists who want to explore how these disciplines connect, TEKHUB Nigeria offers technology learning and practical digital-skills development.

Final thoughts

Apple HomeView brings AI assistance and smart home management into a shared display. Its long-term value will depend on how well its features work together, how reliably it controls compatible devices, and how effectively it protects household information.

For developers, the broader lesson is clear: intelligent interfaces are only as good as the systems, safeguards, and integrations behind them.

What do you think will be the biggest engineering challenge for AI-powered smart home devices: interoperability, privacy, or reliable execution?

AI #IoT #SmartHome #Cybersecurity #WebDev

When a Cyberattack Heats Supermarket Refrigerators: Lessons for Developers and Security Engineers

Tekk NG — Wed, 14 Oct 2026 11:38:53 +0000

Most cybersecurity headlines involve compromised accounts, stolen credentials, ransomware or leaked databases. But what happens when an attacker gains control of a system responsible for keeping food cold?

On October 12, 2026, a cyberattack targeting connected refrigeration systems affected hundreds of French supermarket locations. Some cold-storage temperatures reportedly reached 50°C, with food losses and fires among the reported consequences.

Separately, TFG London, the owner of Phase Eight, Hobbs and Whistles, reported a cyber incident that disrupted some IT systems and online services. Investigations into both incidents were ongoing, and no confirmed connection between them had been established.

For developers and security engineers, these incidents provide a practical lesson: when software controls real-world operations, a security failure can become a physical and financial problem.

1. Operational technology is not just another web application

A typical web application may manage user accounts, content or online transactions. Operational technology (OT), by contrast, monitors or controls physical processes.

Examples include:

Refrigeration and cold-storage controllers.
Industrial automation systems.
Building-management systems.
Energy monitoring and control equipment.
Sensors and actuators connected through industrial networks.

These systems have different availability and safety requirements from many ordinary applications. A patch that is routine for a web server may require careful planning for equipment that must operate continuously.

Security engineers need to understand the process being controlled, not just the software interface.

2. Connectivity expands the attack surface

Remote monitoring and management can improve efficiency, but every unnecessary connection creates potential exposure.

Common risk factors include outdated firmware, exposed management interfaces, weak credentials, excessive privileges and insufficient separation between operational and corporate networks.

A compromised controller may be able to change operating parameters, depending on its permissions and the safeguards around it.

The French incident demonstrates why the security of connected equipment should be considered part of the overall system design.

3. Defence in depth for connected systems

No single security control is sufficient. A more resilient design combines multiple safeguards.

Asset inventory and lifecycle management

Maintain an accurate inventory of controllers, gateways, firmware versions and network connections. Identify unsupported equipment and plan upgrades or replacements.

Network segmentation

Separate operational technology from ordinary office networks and internet-facing services. Restrict communication to what the system actually requires.

Strong access control

Remove default credentials, limit administrative access, use strong authentication where supported and disable unnecessary remote-management services.

Monitoring

Monitor configuration changes, unusual commands, unexpected connections and deviations from normal operating behaviour. Where practical, correlate digital events with physical sensor readings.

Safe recovery

Prepare procedures for isolating compromised systems, switching to safe operating modes and restoring normal service. Test those procedures without putting people or equipment at risk.

4. Security testing must account for physical consequences

For a conventional web application, an unauthorised configuration change may cause incorrect output or expose data. In an operational environment, the same type of weakness could affect temperature, pressure, machinery or other physical conditions.

This changes how teams should assess risk.

Threat modelling should identify which systems can influence physical processes, what the consequences of unauthorised control could be, and how operators can detect and safely respond to abnormal behaviour.

Testing should be authorised and performed in controlled environments. Experiments against live industrial equipment can create safety risks and operational disruption.

5. The separate retail incident offers another lesson

The incident involving Phase Eight, Hobbs and Whistles demonstrates a different dimension of cybersecurity: availability and business continuity.

When online orders or internal systems become unavailable, companies may struggle to process transactions, communicate with customers or fulfil orders.

Developers can help reduce these risks through secure deployment practices, least-privilege access, tested recovery procedures, dependency management and clear incident-response playbooks.

Security is not only about preventing unauthorised access. It is also about ensuring that a service can recover safely when something goes wrong.

6. Why this matters to developers in Nigeria

As businesses digitise retail, logistics, manufacturing and other services, the need for people who understand secure software and connected systems will grow.

Developers do not all need to become industrial security specialists. However, understanding authentication, networking, secure configuration, monitoring and incident response makes it easier to build systems that are resilient from the beginning.

For developers and learners in Nigeria, TEKHUB provides a starting point for exploring practical technology skills and learning pathways relevant to the wider digital ecosystem.

Final thoughts

The French refrigeration attack is a reminder that software does not operate in isolation. It can influence equipment, inventory, services and people's livelihoods.

The separate fashion-retail incident shows that even without physical equipment damage, a cyber incident can interrupt the systems customers and businesses rely on.

For developers and security engineers, the takeaway is straightforward: understand the real-world impact of the systems you build, restrict access appropriately, monitor for abnormal behaviour and plan for safe recovery.

If you were designing security for a connected refrigeration system, what would you prioritise first: network isolation, secure remote access, monitoring or safe manual controls?

Local AI Is Becoming a Hardware Feature: What Developers Should Know About the New Surface and Galaxy Tab Devices

Tekk NG — Tue, 13 Oct 2026 12:54:06 +0000

AI development is no longer limited to cloud APIs and remote GPU clusters. As processors and memory systems improve, more AI workloads can run directly on personal devices.

Two recent hardware announcements illustrate this shift:

Microsoft's Surface Laptop Ultra, designed around Nvidia's RTX Spark platform.
Samsung's Galaxy Tab S12 Ultra and S12+, powered by MediaTek's Dimensity 9500.

Both platforms target demanding workloads, including AI-assisted productivity and creative applications. Microsoft's device is aimed at users who need substantial local computing power, while Samsung's tablets bring AI processing capabilities to a portable form factor.

For developers, the important story is not just the hardware. It is how local compute changes application architecture.

1. Local inference versus cloud inference

A traditional cloud-based AI application typically follows this flow:

The user submits a request.
The application sends data to a remote API.
A cloud service runs the model.
The result returns to the application.

This architecture gives developers access to large models and scalable infrastructure, but it can introduce network latency, service costs and data-handling considerations.

With local inference, a supported model runs directly on the device. This can reduce network dependence and keep certain data local, but the available compute, memory and power budget are limited.

Neither approach is universally better. The right choice depends on the task, the model and the product requirements.

2. What stronger local hardware enables

Devices built around capable CPU, GPU and AI-processing resources can support a wider range of local workloads.

Potential examples include:

Summarising local documents.
Assisting with text and image editing.
Running smaller language models.
Performing selected image-processing tasks.
Prototyping AI features without making every request to a cloud service.

More demanding models may still require remote infrastructure. Model size alone also does not determine quality or speed; quantisation, context length, memory bandwidth and implementation all matter.

3. Hybrid AI architecture is often the practical choice

For production applications, a hybrid approach can offer a useful balance.

For example, an application could use a small local model for quick classification or simple assistance, then call a cloud model for more complex reasoning.

A developer might evaluate the design using metrics such as:

Latency: How long does the user wait?
Memory usage: Can the model fit within the available resources?
Energy use: How much battery does the workload consume?
Privacy: What data leaves the device?
Cost: How many cloud requests can be avoided?
Reliability: What happens when the device is offline?

The architecture should be driven by measured performance and user needs rather than the assumption that either local or cloud AI is always preferable.

4. Hardware capability does not replace software optimisation

A powerful processor will not automatically make an application efficient.

Developers still need to choose appropriate model formats, manage memory, minimise unnecessary computation and test on real devices. They must also consider operating-system support, hardware compatibility and security updates.

For AI-enabled products, testing should cover both functional correctness and resource consumption. A feature that works well on a high-end development machine may perform poorly on a less capable device.

5. What this means for developers building in Africa

Local AI hardware could create opportunities for developers to build more responsive tools, experiment with models and design applications for environments where connectivity is inconsistent.

But the underlying skills remain essential: programming, APIs, data handling, software testing, cloud deployment and cybersecurity.

For developers in Nigeria who want to strengthen those foundations, TEKHUB offers technology learning pathways that can help learners explore practical skills and build towards real-world projects.

Final thoughts

The Surface Laptop Ultra and Galaxy Tab S12 series reflect a broader move towards AI-capable personal devices.

For developers, the opportunity is to design applications that make sensible use of both local and cloud computing—choosing the right execution environment for each task.

If you were building an AI-powered application today, which workloads would you keep on-device, and which would you send to the cloud?

Amazon's Reported $7B Decart Deal: Generative Video Is Only Part of the Story

Tekk NG — Mon, 12 Oct 2026 11:06:06 +0000

AI acquisitions often make headlines because of the size of the proposed deal. But for developers, the more interesting question is usually this:

What technical capability is the buyer trying to acquire?

Amazon is reportedly in talks to acquire Decart for around $7 billion. The potential deal was reported by The Wall Street Journal and Reuters, and negotiations have not yet produced a confirmed final agreement.

Decart is working on generative video and software that helps AI workloads run across different chips. That combination makes the story worth examining from both an application-development and infrastructure perspective.

1. Generative video is moving towards real-time interaction

Decart's Lucy model supports real-time video editing and transformation.

Traditional video-generation workflows often involve submitting a prompt, waiting for a result and then making revisions. Real-time systems aim to make the experience more responsive, allowing changes to happen as a user interacts with the application.

This introduces engineering challenges beyond model quality:

Latency and responsiveness.
Throughput under concurrent usage.
Compute cost per session.
Streaming and delivery architecture.
Consistency and quality of generated output.
User privacy and content safety.

A compelling demonstration is one thing. Delivering a stable, cost-effective production service is another.

2. Why chip-agnostic AI software matters

Decart also develops software that helps AI workloads operate across different processors.

This is important because AI infrastructure is not built around a single universal chip. Cloud providers and hardware vendors use different processors, software stacks and optimisation techniques.

Running a workload efficiently across hardware platforms can require changes to execution, memory management, model serving and performance tuning.

For cloud providers, software that reduces friction between AI applications and their hardware could make their infrastructure more attractive to developers.

Amazon has developed custom AI chips, including Trainium, as part of its AWS strategy. If it acquires Decart, one possible strategic benefit would be gaining technology that complements that hardware ecosystem. The exact integration plan, however, has not been announced.

3. The AI stack is becoming a competitive battleground

A simplified view of an AI product includes several layers:

Product layer: Web apps, APIs and user-facing experiences.
Model layer: Generative models and inference pipelines.
Runtime layer: Serving, scheduling and workload optimisation.
Compute layer: CPUs, GPUs and specialised accelerators.
Infrastructure layer: Networking, storage, cloud services and power.
Operations layer: Monitoring, reliability, security and cost management.

Improvements at one layer can affect the performance and economics of the entire product.

For example, a more efficient inference pipeline may reduce cost per request. Better hardware utilisation may allow a team to serve more users with the same resources. Lower latency can improve the experience of interactive applications.

This is why AI infrastructure software can become strategically valuable even when it is less visible to end users.

4. What developers can learn from this

You don't need access to a large GPU cluster to understand the engineering principles behind this story.

Developers can start by learning how to:

Build applications that integrate AI APIs.
Measure latency, throughput and request costs.
Deploy services reliably in cloud environments.
Work with data pipelines and model-serving endpoints.
Secure user data and manage access permissions.
Compare technical performance with actual business value.

These fundamentals are useful whether you're building an AI-powered web app, a developer tool or a larger production system.

For developers in Nigeria who want to strengthen their foundations, TEKHUB offers a place to explore practical technology learning pathways and skills relevant to the wider software ecosystem.

5. A $7 billion deal is not the same as a $7 billion outcome

The reported valuation is striking, but acquisition price alone does not tell us whether a deal will succeed.

The buyer must integrate the technology, retain key talent, support existing products and generate value that justifies the cost. Technical differentiation must translate into real-world usefulness.

Until negotiations conclude, the acquisition remains a possibility rather than a done deal.

Final thoughts

Amazon's reported interest in Decart highlights two areas developers should watch: interactive generative AI and software that makes AI workloads more efficient across different chips.

The next wave of AI products will depend not only on more capable models, but also on the engineering systems that make them responsive, affordable, secure and reliable.

If you were building an AI platform today, which would you prioritise first: better models, lower inference costs or infrastructure flexibility?

SpaceX's Reported $40B Nvidia Chip Plan: AI Is Also an Infrastructure Problem

Tekk NG — Fri, 09 Oct 2026 12:42:04 +0000

When we talk about AI development, we tend to focus on models, prompts, frameworks and applications.

But deploying AI at scale introduces a different set of engineering challenges: compute capacity, data-centre infrastructure, networking, power consumption, reliability and cost.

A recent report puts this into perspective: SpaceX is seeking around $40 billion in financing to purchase Nvidia AI chips. Reuters reports that the proposed financing could include $10 billion in bank loans and $30 billion in investment-grade debt, with Apollo Global Management expected to lead the deal.

This is a reported financing plan, not a completed transaction. Nevertheless, it raises useful questions about the infrastructure required to build and operate AI systems.

1. Why do AI workloads need specialised hardware?

Training and running modern AI models can involve enormous amounts of parallel computation.

GPUs and other specialised accelerators are designed to handle many of these operations efficiently. In data-centre environments, they work alongside CPUs, memory, high-speed networking and storage.

The result is not simply a collection of powerful chips. It is a coordinated system where the performance of one component can affect the rest.

For example, adding more GPUs does not necessarily deliver a proportional increase in useful output if networking, memory bandwidth, data pipelines or power capacity become bottlenecks.

2. The stack behind an AI service

A simplified view of a production AI system looks something like this:

Application layer: The product, API or user interface.
Model layer: The model and inference or training workloads.
Compute layer: CPUs, GPUs and specialised accelerators.
Platform layer: Containers, orchestration, monitoring and deployment tooling.
Infrastructure layer: Data centres, networking, storage, power and cooling.
Security layer: Identity, access controls, data protection and system monitoring.

Each layer introduces engineering decisions. Scaling an AI service means thinking about the whole system rather than just the model.

3. Compute capacity has a financial cost

SpaceX's reported financing plan highlights another engineering reality: infrastructure decisions are also financial decisions.

Large-scale compute requires capital expenditure, energy, maintenance and people to operate it. Debt financing may accelerate infrastructure deployment, but the resulting capacity must generate enough value to justify the cost.

For teams building AI products, this is a reminder to measure more than model quality. Useful metrics can include:

Cost per request or task.
Latency and throughput.
GPU utilisation.
Energy consumption.
Availability and failure rates.
Cost of serving each customer.
Quality relative to the resources consumed.

A technically impressive system can still be commercially unsustainable if its operating costs are too high.

4. Why this matters to developers

You do not need to build a hyperscale data centre to learn from this story.

Developers can strengthen their understanding of AI systems by learning how APIs, databases, cloud services, deployment pipelines and security controls fit together.

As AI tools become more common, full-stack developers, backend engineers, cloud practitioners, data professionals and cybersecurity specialists will all have roles to play in building reliable applications.

For developers in Nigeria, investing in these foundations is one way to prepare for opportunities across the wider technology ecosystem. Resources and training pathways are available through TEKHUB.

Final thought

The next phase of AI will be shaped by more than model architecture. It will also depend on how efficiently organisations acquire, deploy, secure and operate computing infrastructure.

SpaceX's reported $40 billion Nvidia-chip plan is a striking example of the scale involved.

What do you think will become the biggest bottleneck for AI at scale: chips, energy, networking, cost or the engineering talent needed to connect everything?

The AI Boom Is Really a Hardware Problem Too

Tekk NG — Thu, 08 Oct 2026 12:28:09 +0000

When developers talk about AI, we usually talk about models.

LLMs. Agents. RAG. Inference. Fine-tuning. APIs.

But there's a physical layer underneath all of it:

Silicon.

And two semiconductor developments this week are a useful reminder of how much the AI boom depends on hardware.

TSMC reported record Q3 2026 revenue of NT$1.49 trillion ($46.71 billion), up 50% year over year, with strong AI demand driving the result.

Meanwhile, Wolfspeed announced a conditional financing commitment of up to $1.5 billion to support U.S. silicon-carbide and wide-bandgap semiconductor production.

Let's break down why developers should care.

Your AI API has a physical dependency

Imagine this:

Your application
↓
AI API
↓
Inference infrastructure
↓
GPU / accelerator
↓
Data centre
↓
Semiconductor
↓
Fabrication facility

We usually think about the top of this stack.

The semiconductor industry operates much closer to the bottom.

And when AI demand increases, pressure moves down the stack.

More AI applications → more inference → more compute → more chips.

TSMC is sitting in the middle of this trend

TSMC is one of the world's largest semiconductor manufacturers and supplies major technology companies, including Nvidia and Apple.

Its Q3 revenue reached $46.71 billion, a 50% year-over-year increase.

That is not simply a financial statistic.

It is a signal about how much money is currently flowing into the infrastructure required to support advanced computing.

September revenue alone rose 54.6% year over year.

But AI isn't the only reason semiconductor infrastructure matters

This is where Wolfspeed becomes interesting.

Wolfspeed focuses heavily on silicon carbide (SiC) and wide-bandgap semiconductor technologies.

These aren't primarily about running an LLM.

They are important for power electronics.

Think:

Electric vehicles
Charging systems
Energy infrastructure
Industrial power systems
High-voltage applications

Wolfspeed's announced financing commitment of up to $1.5 billion is intended to strengthen domestic production of silicon-carbide materials and power devices, alongside other wide-bandgap capabilities.

So we have two different semiconductor stories:

TSMC
AI → Advanced computing → Advanced chips

Wolfspeed
Power electronics → SiC → Efficient power devices

Different applications.

Same underlying lesson:

Semiconductors are infrastructure.

What should developers take from this?

First, understand that software doesn't exist independently of hardware.

The performance of an AI application depends partly on the infrastructure underneath it.

This is why concepts such as:

GPU architecture
Memory bandwidth
Inference optimisation
Quantisation
Model size
Latency
Power consumption

are becoming increasingly relevant to software engineers working with AI.

The stack is getting deeper

As developers, we're increasingly expected to understand more of the system.

A frontend developer may need to understand API latency.

A backend engineer may need to understand infrastructure.

An ML engineer needs to understand GPU workloads.

An AI engineer needs to understand inference economics.

A cloud engineer needs to understand compute capacity.

The boundaries between software and infrastructure are becoming less rigid.

For developers and technology learners in Nigeria, practical training environments such as TEKHUB [www.tekhub.ng] provide pathways into software development, AI, data, cybersecurity and other areas that form parts of this broader technology ecosystem.

The bigger takeaway

The AI revolution is often described as a race to build smarter models.

But TSMC's record quarter reminds us that models need machines.

And Wolfspeed's silicon-carbide investment story reminds us that even the machines depend on increasingly sophisticated semiconductor technologies.

So the next time you call an AI API, remember:

There is an enormous physical infrastructure behind that request.

The AI future is being written in software, but it is being manufactured in silicon.

MTN Nigeria Just Crossed 100 Million Subscribers. What Does That Mean for Developers?

Tekk NG — Wed, 30 Sep 2026 11:24:35 +0000

When a telecom company reaches 100 million subscribers, it is easy to treat it as a business headline.

But for developers, product builders and technology professionals, there is another way to look at the number.

100 million active subscriptions = an enormous connected user base.

MTN Nigeria recorded 100,859,587 active subscriptions in July 2026, according to Nigerian Communications Commission data. That made it the first Nigerian telecom operator to cross the 100-million mark. Nigeria's total active mobile subscriptions stood at 195.11 million.

That scale has implications for the products developers build.

Mobile connectivity is now part of the application stack

Think about a typical Nigerian digital product.

The network isn't the application itself.

But without it, the application can't reach the user.

And users aren't just making calls anymore

Mobile networks increasingly carry:

API requests
Video
Cloud traffic
Financial transactions
Social media
Streaming
Remote-work applications
Online learning
E-commerce activity

MTN reported 55.7 million active data users in the first half of 2026, while average monthly data consumption was reported at 14.8GB per user.

That has consequences for developers.

Build for real network conditions

A Nigerian application doesn't exist in a vacuum.

Users may have:

Different network speeds
Limited data bundles
Older smartphones
Intermittent connectivity
High latency
Limited storage
Different levels of digital literacy

That makes performance an important part of product development.

Developers should think about:

Smaller payloads

Don't send 5MB when 500KB will do.

Image optimisation

Compress images and use appropriate formats.

Caching

Don't repeatedly download information that hasn't changed.

Graceful failures

Your application shouldn't completely collapse because the network disappears for 10 seconds.

Progressive loading

Let users see useful content before everything has finished loading.

API efficiency

Reduce unnecessary requests and avoid expensive operations where possible.

These aren't simply “nice-to-have” optimisations.

They can directly affect the experience of users accessing your application over mobile networks.

More connectivity also means more opportunity

A larger connected population creates a larger potential audience for Nigerian developers.

Fintech.

EdTech.

HealthTech.

E-commerce.

Productivity tools.

Social platforms.

AI applications.

SaaS.

The infrastructure supporting the users is expanding, while the opportunity to build for those users expands alongside it.

Of course, connectivity alone doesn't guarantee that a product will succeed.

But it provides the infrastructure through which digital products can reach people.

Developers need to think beyond the interface

This is also why learning technology isn't only about writing code.

A developer building for a large audience needs to understand:

Performance + accessibility + security + APIs + infrastructure + user behaviour.

For Nigerians developing these skills, TEKHUB is one example of a local technology-training ecosystem where learners can develop practical skills across software development, cybersecurity and other digital disciplines.

MTN's 100-million milestone isn't necessarily a developer statistic.

But it tells developers something important:

Nigeria already has an enormous connected audience.

The opportunity is to build products that are actually worth connecting to.

Ethical Hacking Is Only One Layer of Modern Cybersecurity

Tekk NG — Tue, 29 Sep 2026 13:00:04 +0000

If you're learning cybersecurity, there's a good chance you've encountered the usual roadmap:

Networking → Linux → Python → Kali Linux → OWASP → Penetration Testing

That's a useful starting point.

But there's another part of cybersecurity that doesn't always get enough attention:

What happens after you find the vulnerability?

That's where things get interesting.

Recent cybersecurity discussions in Nigeria increasingly connect technical security with data protection, cloud security, governance and cyber resilience. A Nigeria Computer Society workshop, for example, covered access control, breach prevention, third-party and cloud security risks, privacy-by-design and digital governance alongside cybersecurity.

Nigeria's broader cloud policy direction is also bringing digital sovereignty and infrastructure governance into focus.

The penetration test isn't the finish line

Suppose you're testing an application and discover an insecure API endpoint.

You report it.

Great.

But now the organisation needs to understand:

What can an attacker access?
Is authentication being bypassed?
Does the endpoint expose personal data?
Is the vulnerability present in production?
Are logs capturing exploitation attempts?
What is the remediation priority?
Could the same issue exist elsewhere?

Suddenly, a technical finding has become a risk-management problem.

That's cybersecurity.

Cloud makes this even more interesting

Modern applications are rarely sitting on one physical server.

For example:

Who can access the database?
Are cloud credentials protected?
Are permissions excessive?
Is sensitive data encrypted?
Are logs being monitored?
Can compromised credentials be revoked quickly?
What happens if the cloud environment is unavailable?

This is why cloud security and ethical hacking increasingly overlap.

Then comes data sovereignty

Another question is:

Where is the data actually being stored and processed?

Nigeria's National Digital Cloud Policy places digital sovereignty and safeguards for government and regulated data within its cloud infrastructure framework.

For developers and security engineers, this means infrastructure decisions can have implications beyond performance and cost.

They can also affect:

Security + privacy + compliance + governance.

What should aspiring ethical hackers learn?

Don't abandon the technical fundamentals.

Learn them properly.

But consider expanding your roadmap:

Networking

TCP/IP, DNS, HTTP, ports, routing and network security.

Linux

Command line skills, permissions, processes and system administration.

Web security

OWASP Top 10, authentication, authorisation, APIs and common web vulnerabilities.

Penetration testing

Learn how to identify, validate and responsibly report vulnerabilities.

Cloud security

Understand IAM, permissions, secrets, logging and common cloud misconfigurations.

Security operations

Learn how organisations detect, investigate and respond to threats.

Governance and compliance

Understand why security controls exist and how organisations demonstrate that those controls are working.

This is the direction cybersecurity is moving toward: technical depth combined with a better understanding of the environment around the technology.

For learners in Nigeria, TEKHUB [www.tekhub.ng] includes Ethical Hacking + Cyber-Security in its academy, with practical areas such as Kali Linux, OWASP, penetration testing, vulnerability assessment and network security.

The goal of ethical hacking isn't to become better at breaking things.

It's to become better at understanding how things can break—and helping organisations prevent it.

AI Adoption in Nigerian SMEs: The Gap Between “Using AI” and Actually Integrating It

Tekk NG — Mon, 28 Sep 2026 10:47:21 +0000

There's a common assumption in discussions about AI adoption:

If people are using AI tools, then AI adoption is happening.

Not necessarily.

A business owner using ChatGPT to write a caption is very different from a business integrating AI into customer support, reporting, operations or decision-making.

That's the gap highlighted by the new AI Skills for Business Programme from HP and UpSkill Universe.

The programme aims to train 55,000 SMEs across Nigeria and South Africa using HP LIFE, with additional live training and an intensive bootcamp for selected participants.

The numbers

UpSkill Universe says a recent study found:

Nearly 70% of SMEs remain in the early stages of AI adoption.
Only 9% have embedded AI into day-to-day operations, strategy and decision-making.

These figures are from a study cited by the programme organisers, so they shouldn't automatically be treated as a universal benchmark for every Nigerian SME.

But the distinction they highlight is useful.

AI usage
Business owner
↓
Uses ChatGPT occasionally
↓
Generates content / ideas
AI integration
Business process
↓
AI-assisted workflow
↓
Human review
↓
Business decision/action
↓
Measurable result

The second requires considerably more planning.

Why developers should care?

Developers often focus on the technical side:

APIs
Models
Prompt engineering
RAG
Agents
Automation
Data pipelines

But SMEs don't necessarily start with those technologies.

They start with a problem.

For example:

“I spend four hours every week preparing reports.”

That's a business problem.

AI might become part of the solution.

Another business might say:

“We receive too many repetitive customer questions.”

That's another problem that could potentially be addressed with automation or AI-assisted customer support.

The technology comes after understanding the workflow.

What the programme is doing

The HP-UpSkill programme has a three-level structure.

55,000 SMEs

Participants will be onboarded to HP LIFE and expected to complete at least three AI courses.

5,000 SMEs

Five thousand participants will receive live 90-minute facilitator-led training focused on real business challenges.

Each participant is expected to leave with an AI tool to test within 30 days.

50 SMEs

The top 50 performers will move into a Build Your Business with AI Bootcamp involving business audits, HP mentorship and 90-day action plans.

The programme also aims for 40% participation from youth and women-led businesses.

AI implementation isn't just an engineering problem

Suppose a developer builds a chatbot for an SME.

Technically, it works.

But then what?

Who maintains it?

Where does its knowledge come from?

What happens when the model gives a wrong answer?

What data can it access?

Who reviews sensitive outputs?

When should a human take over?

How do you measure whether it is actually saving the business time or money?

Those questions are part of AI adoption.

A working AI feature isn't automatically a useful AI system.

The skills layer

This is why business-focused AI training is becoming increasingly relevant.

Business owners don't necessarily need to become machine-learning engineers.

They do, however, need enough understanding to identify suitable use cases, evaluate AI tools, manage risks and integrate the technology into their workflows.

Developers also benefit from understanding the business problems behind the systems they build.

Nigeria's wider technology-training ecosystem includes organisations such as TEKHUB, which provides technology and digital-skills training.

The interesting part: moving from courses to experiments

The programme's 30-day implementation component is particularly relevant.

Learning about an AI tool is one thing.

Actually testing it inside a business creates feedback.

The business can ask:

Did it save time?
Did it improve accuracy?
Did customers respond better?
Did it reduce repetitive work?
Did it create a new problem?
Is the cost justified?

Those are much more useful questions than simply asking whether a company “uses AI.”

Final thought:

The AI adoption conversation shouldn't only be about how many businesses have access to AI.

It should also be about how many businesses know what to do with it.

The difference between experimentation and integration may become one of the defining digital-skills challenges for African SMEs over the next few years.

Cybercrime Isn't Always Hacking: What the Black Axe Extradition Case Teaches Us About Social Engineering

Tekk NG — Fri, 25 Sep 2026 12:57:14 +0000

When people hear “cybercrime,” they often imagine someone exploiting a vulnerability in a server.

But what if the vulnerability is a human relationship?

That's the interesting cybersecurity angle behind a recent international criminal case involving six Nigerian nationals extradited from South Africa to the United States.

On September 11, 2026, South African authorities handed the six men over to US authorities. They are accused of being members of the Black Axe organised crime network and face US charges including wire fraud and money laundering.

The allegations involve online romance scams, advance-fee schemes and other forms of internet fraud.

The alleged attack surface: trust

According to the US Department of Justice, the alleged schemes operated from Cape Town between 2011 and 2021.

The defendants allegedly used:

Social media
Online dating websites
VoIP phone numbers
False identities
Fabricated stories

to communicate with victims in the United States.

The alleged romance-scam narratives reportedly involved supposed romantic partners travelling to South Africa for work and later claiming to experience emergencies requiring money or valuable items.

That's social engineering.

The attacker doesn't necessarily need to exploit a software vulnerability.

They exploit a person's trust.

The alleged scale

South Africa's Hawks said the six men allegedly targeted more than 100 women in the United States.

The alleged losses were reported at more than R100 million, or roughly $6.2 million according to Reuters.

Some alleged victims were pensioners and businesspeople.

The six were originally arrested in South Africa in 2021.

After the extradition process, they were transferred to US custody on September 11, 2026.

The cybersecurity lesson:

Consider two attack models.

Model 1: Technical exploitation
Attacker
↓
Find vulnerability
↓
Exploit system
↓
Gain access
↓
Steal/manipulate data

Model 2: Social engineering

Attacker
↓
Create fake identity
↓
Establish trust
↓
Manipulate victim
↓
Victim takes action
↓
Money/data/access is compromised

The second model doesn't necessarily require sophisticated malware.

The human being becomes the attack surface.

Why developers should care

If you're building an application, security isn't just about writing secure code.

You also need to consider how your users can be manipulated.

For example:

Can an attacker impersonate another user?
Can someone easily change account details?
Are financial transactions reversible?
Can users report suspicious activity?
Are warnings shown before high-risk actions?
Are authentication and recovery flows secure?
Are sensitive actions logged?
Can an attacker abuse your messaging system?

Security UX matters.

A technically secure backend can still support an unsafe user experience.

Cybersecurity education goes beyond tools

Learning cybersecurity isn't simply learning how to use a penetration-testing tool.

It also involves understanding:

Social engineering
Threat modelling
Digital forensics
Identity management
Incident response
Fraud patterns
Security awareness
Risk management

For people building careers in cybersecurity and other technology fields, Nigerian training organisations such as TEKHUB are part of the wider skills ecosystem; its technology programmes can be explored at http://www.tekhub.ng.

International cooperation is another part of the story

The case also demonstrates the international nature of cyber-enabled crime.

South African authorities conducted the arrests and coordinated the extradition.

INTERPOL South Africa participated in the handover.

The FBI and US Secret Service received the suspects.

US prosecutors are pursuing the case.

The internet doesn't respect national borders, so investigating digital crime often requires cooperation across them.

One important distinction

The defendants have been charged, not convicted.

The US Department of Justice explicitly says that the charges and allegations are accusations and that the defendants are presumed innocent unless proven guilty.

That's an important distinction when discussing cybersecurity cases.

Final takeaway

The Black Axe case is a reminder that cybersecurity isn't always about defeating code.

Sometimes the attack begins with a message:

“Hello, I think you're interesting.”

The technology provides the channel.

The social engineering provides the manipulation.

And the victim's trust becomes the target.

NCC Zero-Rated Education Access in Nigeria: Does This Make Online Courses Free?

Tekk NG — Thu, 24 Sep 2026 11:55:35 +0000

If you're learning to code in Nigeria, you probably already know that “free course” doesn't necessarily mean free learning.

The course might cost ₦0.

The internet connection doesn't.

That's why the Nigerian Communications Commission's new Zero-Rated Data Access to Educational Platforms and Content Initiative is worth paying attention to.

The NCC launched the initiative in collaboration with the Federal Ministry of Education to reduce data-cost barriers to digital education.

But there's an important technical distinction:

Zero-rated educational access is not the same thing as free internet.

What is zero-rating?

Imagine you have 2GB of normal mobile data.

You visit an approved educational platform under the new programme.

The data consumed while accessing that platform doesn't reduce your normal 2GB balance.

That's the basic idea behind zero-rating.

The initial programme provides eligible users with up to 100MB per day for accessing designated educational platforms. It is scheduled to commence on October 1, 2026.

So can I take any online course for free?

No.

And this is probably the most important thing developers and other digital learners need to understand.

The NCC isn't saying:

IF website = educational
THEN data = free

Instead, the programme works around approved/whitelisted platforms.

The NCC says the participating ecosystem includes categories such as:

Learning Management Systems
Digital libraries
Educational repositories
Teacher-development platforms
Technical and vocational training platforms

Platforms have to go through the programme's approval process.

A zero-rated programme covering one approved platform doesn't automatically make everything in that workflow free.

This distinction matters because modern technical learning is rarely contained inside a single website.

A learner may need documentation, repositories, videos, APIs, forums and course material to complete one project.

But there is a promising part,

The NCC's framework specifically includes technical and vocational training platforms.

That's relevant to the growing number of Nigerians learning software development, cybersecurity, data, cloud computing and other technical skills.

For example, Nigerian technology-training organisations such as TEKHUB operate within this broader digital-skills ecosystem. You can explore TEKHUB's technology training resources at http://www.tekhub.ng.

Whether any specific platform qualifies, however, depends on its participation and approval under the programme.

What does 100MB actually get you?

It depends on the content.

Text is relatively cheap in bandwidth terms.

Video isn't.

A developer reading documentation, PDFs or text lessons may be able to learn for considerably longer than someone streaming video lectures.

The NCC estimates that 100MB can support different amounts of educational activity depending on the format, and says the cap is partly intended to keep the programme sustainable while usage is assessed.

That makes optimisation interesting from a technical perspective too.

Educational platforms participating in the initiative are being encouraged to reduce unnecessary data consumption, including optimising how content is delivered.

For developers building educational products, that is a useful reminder:

Performance isn't only about how quickly a website loads.

For some users, bandwidth efficiency determines whether they can use the product at all.

What is actually free?

Think of the initiative as:

free access to approved educational destinations within the programme's allowance

rather than:

free access to the internet.

That's a major difference.

If you're taking a course on an approved platform, the relevant access may be zero-rated.

If you're downloading something from an unrelated website, the normal data rules may still apply.

What happens next?

The NCC says platforms will be monitored based on factors such as usage, data consumption, service availability, consumer complaints, network performance and regulatory compliance. More platforms can also be incorporated over time.

That means the initial launch isn't necessarily the final shape of the programme.

For Nigerian developers and students, the interesting question is therefore not simply:

“Is internet free now?”

It's:

“Which educational platforms are covered, who qualifies, and how useful can 100MB a day be for real digital learning?”

Those are the questions worth watching as implementation begins.

CyberSecurityNigeria #DigitalSkillsNigeria #NITDA #NCC