91ÖÆƬ³§

AWS Summit 2025, Washington DC – Focus on AI Workloads

On June 10-11, 2025, AWS held its AWS Summit 2025, in Washington, DC. This annual event had a strong focus on the public sector with an emphasis on cloud computing, AI, and national security.

Key themes of this conference included: Agentic AI, AI sovereignty, and how cybersecurity has been transformed with the use of AI.

ANALYST ANALYSIS

Key announcements at this conference include:

• The AWS Secret-West Region to be launched by the end of 2025 to support workloads up to the secret classification level. This marks the second AWS region for this kind of workload – in 2017, AWS launched its first AWS secret region.
• Amazon to invest $20 billion in Pennsylvania and $10 billion in North Carolina. The investments have been cited to support growing demand for generative and agentic AI.

These investments demonstrate how important AI infrastructure is, especially to support secure and compliant AI-powered workloads for national security, government agencies, education, and healthcare.

Microsoft 91ÖÆƬ³§ 2025 Conference – Focus on Developers and AI

On May 19-22, 2025, Microsoft held its premier annual developer conference, Microsoft 91ÖÆƬ³§, in Seattle.

Windows AI Foundry was announced as Microsoft's strategic platform for developing AI applications to run locally on Windows devices, particularly for Copilot+ PCs (which run on NPU-based processors). This platform allows developers to access both open-source and closed-source AI models, optimize inference workloads, and customize models.

ANALYST ANALYSIS

Highlights of this conference focused on Microsoft's perspective on the AI agent ecosystem, which included:

• The "Open Agentic Web" vision, which describes future AI workloads being driven by AI agents. It provides support for protocols like Model Context Protocol and Agent2Agent to enable the management and orchestration of AI agents.
• The evolution of copilots to agents. One example of this would be GitHub Copilot's evolution into a coding agent.

It was no surprise that the focus of the conference was on how Microsoft wants to be viewed as the development platform of choice for AI.

DeepSeek-R1-0528 now available

On May 28, 2025, DeepSeek-R1-0528 was made available. This is an update to their DeepSeek R1 reasoning model.

Highlights of this update include improvements to their reasoning and inference capabilities, reducing the probability of producing hallucinations, improved web search, and its coding capabilities.

ANALYST ANALYSIS

The focus of the update includes the improvement of its reasoning and inference capabilities, where it can better handle multi-step, complex tasks. It is comparable to OpenAI's o3 and Google Gemini 2.5 with respect to performance. The code generation capabilities has improved, and its performance is just behind OpenAI o4-mini and o3.

The closed source models continue to deliver faster response times for inference workloads and provides larger context windows.

DeepSeek continues to be the industry leader in providing an open weight model. It outperforms several of the closed source models and continues to provide highly competitive pricing with its open-source pricing format.

Google I/O 2025 Conference – AI Software Update

On May 20-21, 2025, Google held their I/O 2025 conference in Mountain View, California.
The conference highlighted the improvements and increased usage of its Gemini AI model by their own products and by the customers invested in Google-based solutions.
Naturally, there was a strong emphasis on AI agents and its future. Announcements focused on Agentic AI included:

• Google search updated with AI mode
• Jules, a new AI coding assistant
• Gemini Canvas, a code assistant for non-developers
• Project Mariner, enabling 10 agents to operate simultaneously
• Updated agent development kits for Python and Java
• Improvements to its Agent2Agent protocol

ANALYST ANALYSIS

Google's latest conference delivered an update on its AI investments and its latest AI offerings. Some of the announcement highlights include:
Gemini Updates

• New Deep Think Mode reasoning mode for complex, multi-step use cases.
• New Native Audio support, enabling models to "hear" and "speak" with greater understanding and supporting 24 languages.
• LearnLM Integration, a set of fine-tuned models for learning

new multimodal models.

• Veo 3, their latest video generation model with synchronized sound.
• Imagen 4, an update to Google's image generation model with higher quality, suitable for creating marketing collateral.

New Specialized Models

• MedGemma, an update to their multimodal model for the healthcare industry.

Google went all out to demonstrate how pervasive AI across their portfolio of products, from consumer-based to enterprise-based software. Its key message was that AI is deeply integrated into every facet of Google's products and services, with Gemini serving as the foundational intelligence layer.

Apple Execs Defend Siri Delays, AI Plan, and Apple Intelligence

On June 10, 2025, Joanna Stern,

senior personal technology columnist for the Wall Street Journal, interviewed Apple SVP Software Engineering Chief Craig Federighi and SVP Worldwide Marketing Head Greg Joswiak at the recent Apple WWDC 2025 conference in Cupertino.

By the end of the interview, no dates were committed to for an AI-enhanced version of Siri, nor any reasons why anybody would use Apple Intelligence.

ANALYST ANALYSIS

Last year, Apple announced a smarter AI-driven Siri, but it has still not been able to deliver an AI-driven Siri citing their preliminary product did not meet their quality and reliability expectations. The journalist proposed that Siri should be better than, or at least comparable to, other AI assistants today. To compound this, Apple continued to market its upcoming AI capabilities and recently, several class action lawsuits have been launched against Apple regarding their alleged lack of AI capabilities in Siri and the Apple Intelligence features advertised for the iPhone 16.

Unfortunately for Apple, this demonstrates how important it is to develop the right AI strategy for the organization. And despite being one of the most valuable companies in the world, Apple is not immune for making errors in their AI strategy and execution of their AI roadmap.