At Knowledge Architecture, we are deeply committed to delivering innovative solutions while being mindful of the environmental impact of AI.
Currently, there isn’t a standard way to measure the environmental impact (specifically, energy and water usage) of using AI. There aren’t industry-wide protocols, regulations, or a shared vocabulary. The companies that build the foundational models—OpenAI, Anthropic, Google, and others—are only just beginning to share data, and when they do, they’re often partial, lacking context, and hard to verify.
Still, these questions matter. In this article, we’ll share what we do know at this time. We’ll continue to update it as we learn more.
How Much Energy and Water Does Synthesis AI Search Use?
In June 2025, Sam Altman, the CEO of OpenAI, published what may be the first widely circulated estimate of inference energy and water usage for ChatGPT. He reported that an average query consumes about 0.34 watt-hours, which is equivalent to 0.00034 kilowatt-hours (kWh) of electricity, and 0.000085 gallons of water. That means each query uses about what a high-efficiency lightbulb uses in a few minutes, and a fraction of a teaspoon of water.
The average AEC firm running Synthesis AI Search currently executes about 2,000 queries per month.
According to OpenAI’s estimates, this would consume approximately:
- .68 kWh of energy
- .17 gallons of water
Which is roughly equivalent to:
- driving an efficient EV for 2.8 miles 🚘
- taking a 5-second shower 🚿
What We Do and Don’t Know
The estimate from OpenAI likely only includes the power used during inference, which is the term for running queries to generate a response. It may not account for the rest of the system: cooling, networking, storage, or other overhead. And OpenAI hasn’t said whether that number reflects short prompts, long ones, or some average across different modes and models.
What we do know is that AI systems require more energy than traditional software. When you use AI Search in Synthesis, you are asking a model to do a meaningful amount of work. That work gets done in Microsoft’s Azure infrastructure, not OpenAI’s, since Synthesis uses the Azure OpenAI Service. Azure data centers are among the most efficient in the world, and Microsoft has made strong commitments to being carbon negative and water positive by 2030. The actual, full footprint of any individual query remains hard to pin down.
This is an emerging field. AI’s environmental impact isn’t well understood yet. But we believe it will be—and we believe it should be. Just as appliances have Energy Star labels and cars have fuel efficiency ratings, we hope that AI systems will soon publish standardized, transparent data about their energy and water use. That kind of visibility will make it easier for all of us—users, developers, platform providers—to make better, more responsible choices.
Until then, we’re sharing what we know, what we don’t, and designing Synthesis in a way that gives you options. Because even if the numbers are small, they’re not zero. And even if the systems are opaque, the questions are worth asking.
How Knowledge Architecture Optimizes AI Usage for Synthesis AI Search
AI inference, the process of using AI models to generate insights, can be energy intensive. To address this, we’ve designed our AI infrastructure with efficiency in mind:
- Using Low-Energy Models by Default:
We use smaller, energy-efficient models, like "GPT 4.1 mini," to handle the majority of queries. These models consume significantly less energy while delivering fast and accurate results. - Scaling Up Only When Necessary:
For more complex queries, we may layer in more powerful models in the future, but only if required for a specific query. This ensures we minimize energy while maintaining a high-quality user experience. - Continuous Monitoring and Optimization:
Our team evaluates new AI models to identify opportunities to further improve efficiency and reduce compute requirements.
User and Firm-Level Choice for Default Search Experience
Firms and users can choose how and when to use Synthesis AI Search. Some users and firms will want AI on by default. Others may prefer to begin with Classic Search and opt in to AI-generated summaries as needed.
This approach gives firms more control over when and how AI is used, both for user experience and sustainability.
Leveraging Sustainable Infrastructure
We partner with Microsoft to power our AI solutions, leveraging their state-of-the-art data centers. Microsoft has made bold commitments to sustainability, including:
- Carbon Negative by 2030: Microsoft is working to remove more carbon than they emit by 2030.
- Water Positive by 2030: Microsoft aims to be water positive by 2030 by reducing its water use, sourcing alternatives, and investing in replenishment and access projects.
- Sustainable Data Center Innovations: From using renewable energy to innovative designs and materials, Microsoft continuously reduces the carbon footprint of its infrastructure.
These efforts ensure that the compute resources we use are sourced from facilities actively striving to minimize environmental impact.
Our Philosophy on Sustainable AI
We believe that sustainability is about smart choices:
- On-Demand Compute Power: We scale our compute resources dynamically, using higher-energy models only for the most complex tasks.
- Data-Driven Decisions: By analyzing user interactions, we ensure our AI solutions are optimized to meet your needs without unnecessary energy expenditure.
- Transparency: We are committed to providing insights into how our AI operates and how it aligns with our sustainability goals.
- Balancing Innovation and Choice: We believe in giving users and firms meaningful choices about when and how to use AI. That’s why we’re building features—like configuring Default Search Tabs—that put control in your hands. As AI becomes more deeply embedded in modern software, some future features may require AI. When that happens, we’ll remain committed to transparency and optimizing Synthesis for both user experience and sustainability.
Looking Ahead
Sustainability is an ongoing journey, not a destination. While we are proud of our efforts to date, we remain committed to continuous improvement. As we advance our AI capabilities, we will explore additional strategies to minimize energy use and reduce our environmental footprint.
Please send any questions and/or feedback to support@knowledge-architecture.com.