Meta’s $14.3B AI Bet & Corning’s GenAI Boost Signal Big Tech Growth Surge

U.S. technology leaders are doubling down on artificial intelligence as a growth engine.

On Wednesday, Meta CEO Mark Zuckerberg will face investor scrutiny over a $14.3 billion investment in Scale AI and a high-profile hiring spree—moves prompted by a misstep in copying Chinese MoE techniques and sluggish uptake of Llama 4. At the same time, Corning CEO Wendell Weeks told CNBC that generative AI demand underpinned the glassmaker's latest quarterly results. Meanwhile, Wedbush's Dan Ives forecasts AI-driven "strong results" from Microsoft, Meta, Apple and Amazon in the second half of 2025, with each expected to post double-digit revenue growth.

Adobe has introduced three new Firefly-powered tools in Photoshop's beta builds to accelerate common editing tasks.

• Generative Upscale sharpens and enlarges images up to eight megapixels without sacrificing clarity.

• Harmonize (formerly Project Perfect Blend) automatically matches color, lighting, shadows and tone when you add objects to a scene.

• An improved Remove tool now focuses solely on excising unwanted elements and filling gaps with realistic background textures.

These AI-driven features are available now in Photoshop beta on desktop and web, with Harmonize in early access on iOS. A new Projects asset manager and GenAI Model Picker round out the release.

Autonomous AI agents are breaking out of the lab and into the enterprise. In late July, Writer, a San Francisco–based startup valued at $1.9 billion, rolled out its Action Agent, capable of independently executing complex, multi-step workflows—browsing sites, analyzing data, writing code and updating systems—across 600+ platforms.
Meanwhile, Xebia has made its Agentic AI solutions discoverable and deployable via the new AWS Marketplace AI Agents and Tools category, accelerating procurement for GenAI-powered search, orchestration and governance.
On the standards front, Cisco's AGNTCY project, now stewarded by the Linux Foundation and backed by Dell, Google Cloud and Oracle, aims to forge an "Internet of Agents" with open protocols for discovery, secure messaging and auditability.
This convergence of advanced capabilities, marketplace access and open standards signals a tipping point for agentic AI in U.S. enterprises.

Attackers are now wielding AI chatbots to scale ransomware negotiations, according to Picus Security research. Global Group's chatbot automates communication, ups demands to over $1 million, and applies psychological pressure—while human operators still set deadlines and oversee leaks.

Defenders are fighting back. At Black Hat USA 2025, Darwinium launched two agentic AI tools:

• Beagle: autonomously red-teams customer journeys to expose weak spots


• Copilot: conversational assistant that generates detection rules and quantifies trade-offs

Together, these innovations aim to match AI-driven attacks with AI-powered defenses.

Enfabrica Corporation on July 29 unveiled EMFASYS, the industry's first Ethernet-based AI memory fabric, in Mountain View, California. The appliance marries RDMA-over-Ethernet networking with parallel CXL DDR5 channels to deliver elastic memory bandwidth and capacity to any GPU server at low, bounded latency.

By offloading high-cost HBM to pooled DDR5 DRAM, EMFASYS helps large-scale LLM inference workloads extract higher GPU and memory utilization, targeting a 50% reduction in cost per token. The system is now sampling with customers alongside Enfabrica's 3.2 Tbps ACF-S SuperNIC chip.