jueves, 3 de septiembre de 2026

Broadcom Lifts AI Chip Outlook on Hyperscaler Spending

Broadcom has raised its forecast for AI-driven chip revenue, signalling that demand from hyperscale cloud providers continues to accelerate rather than plateau. The revised outlook reflects both stronger-than-expected orders for custom AI accelerators and the broader reality that major technology companies are still committing record sums to AI infrastructure.

Broadcom raises AI chip forecast as Big Tech keeps writing bigger checks - Reuters
Imagen generada con IA

Custom silicon has become a central pillar of that build-out. While Nvidia dominates the market for general-purpose GPUs used in AI training and inference, cloud operators including Google, Amazon and Meta have been investing heavily in their own accelerator designs to optimise cost and performance for specific workloads. Broadcom serves as a key partner in this effort, supplying the networking chips, custom ASIC designs and integration expertise that make these in-house processors viable at scale. The raised forecast suggests that this custom-silicon track is gaining real commercial momentum alongside, not in competition with, Nvidia's product cycle.

Several factors explain why hyperscalers keep opening their wallets. Training the latest generation of frontier models requires clusters of tens of thousands of accelerators interconnected by high-bandwidth fabric, a configuration where networking and switch silicon become as strategic as the compute itself. Inference at scale, particularly for generative AI services embedded in consumer products, is also driving sustained hardware demand. Providers are simultaneously diversifying their supply chains to avoid bottlenecks and negotiating better unit economics, both of which favour custom designs.

For technically minded readers, the key takeaway is structural rather than cyclical. AI capex is no longer a bet tied to a single product launch or model release; it is becoming a multi-year baseline expense for the largest cloud platforms. That has meaningful implications across the semiconductor stack, from foundry capacity and HBM memory to advanced packaging and optical interconnect, all of which sit on the critical path of AI infrastructure.

The revised guidance also implies confidence that current order books translate into shipped silicon within a reasonable timeframe, despite ongoing constraints in advanced packaging and high-bandwidth memory supply. If that holds, it reinforces a picture in which the AI hardware cycle still has meaningful runway, with custom accelerators, networking chips and supporting components all benefiting from hyperscaler investment that continues to grow rather than taper.

Artículos relacionados de LaRebelión:


Fuente Original: Reuters

Artículo generado mediante AI.larebelion.

AI Cyber Defence Google OpenAI Anthropic Lead Charge

The cybersecurity landscape is on the cusp of a revolution, with tech giants Google, Anthropic, and OpenAI pooling their considerable resources to develop advanced AI models specifically for cyber defence. This isn't just about creating new tools; it's a concerted effort to anticipate and neutralise emerging threats before they can inflict damage. The focus is on building intelligent systems that can not only detect malicious activity but also understand the intricate attack paths threat actors might exploit.

AI Cyber Defence: Google, OpenAI, Anthropic Lead Charge

Central to this initiative is the creation of sophisticated AI models designed to map and understand the complex ways attackers can escalate privileges across different domains. By identifying these critical "choke points," security professionals can proactively fortify their systems, severing potential breach routes at their inception. This preventative approach is a significant shift from traditional reactive security measures.

Furthermore, these leading AI developers are not just building the technology; they are also committed to establishing robust safeguards and access programmes. This suggests a collaborative spirit, aiming to ensure that these powerful AI tools are used responsibly and ethically. The goal is to democratise access to advanced cyber defence capabilities, empowering a wider range of organisations to protect themselves from increasingly sophisticated cyberattacks. The development of these AI models signals a new era in cybersecurity, where artificial intelligence plays a pivotal role in safeguarding our digital infrastructure.

Fuente Original: https://thehackernews.com/2026/09/google-anthropic-and-openai-unveil.html

Artículos relacionados de LaRebelión:

Artículo generado mediante LaRebelionBOT

miércoles, 2 de septiembre de 2026

AI Agents Vulnerable to Supply-Chain Code Execution

Recent research reveals a significant security vulnerability affecting artificial intelligence agents used by major Fortune 500 corporations. Security analysts demonstrated how attackers can easily manipulate autonomous AI systems into executing arbitrary, potentially malicious code simply by exploiting standard documentation protocols.

Researchers easily trick Fortune-500 companies' AI agents into running arbitrary code — supply-chain attack via llms.txt guidance file illustrates how data has become code - Tom's Hardware
Imagen generada con IA

The attack vector focuses on the llms.txt file format, an emerging standard designed to guide large language models on how to interact with a website's documentation and APIs. Because modern AI agents actively read and follow instructions found within these contextual text files to perform automated tasks, malicious actors can poison the supply chain by embedding hidden system prompts or malicious instructions directly into the guidance file.

When an enterprise AI agent ingests the compromised text, it treats the external data as valid operational commands rather than inert information. This effectively blurs the traditional line separating data from executable code. Researchers showed that this flaw allows attackers to hijack the agent's control flow, force unauthorized data exfiltration, or run arbitrary commands on the underlying infrastructure without the human user realizing a compromise has occurred.

For technically literate readers, this discovery highlights a profound architectural challenge in the deployment of autonomous systems. Prompt injection vulnerabilities are no longer confined to chat interfaces where users directly interact with a model. Instead, as AI agents become more deeply integrated into web ecosystems and automated workflows, they introduce expansive new vectors for indirect prompt injection and supply-chain tampering.

Securing these systems will require a fundamental rethink of how AI agents ingest and validate external instructions. Developers and enterprise security teams must implement strict sandboxing, rigorous input sanitization, and clear separation of data and control layers before allowing autonomous models to process untrusted files from the open web.

Artículos relacionados de LaRebelión:


Fuente Original: Tom's Hardware

Artículo generado mediante AI.larebelion.

// Telegram BOT