According to a ranking by TRG Datacenters, which reveals the top U.S. cities for tech jobs, tech workers in Los Angeles earn an average of $146,550, with more than 359,000 jobs currently available, the highest in the country.

Reviewing the data, the U.S. tech map is no longer dominated by the usual suspects. Yes, San Francisco still commands the highest salaries, $150,387 on average, but it lags in startup formation and job numbers. 

Los Angeles, in contrast, is building a bigger base, nearly 25,000 startups launched annually and a growing share of remote roles (19.7%).

Among the top U.S. cities for tech jobs, New York follows close behind L.A. It’s the country’s leader in startup creation, nearly 37,000 new tech businesses per year, and offers an average salary of $148,138. 

Meanwhile, Seattle is becoming the capital of remote tech work. Over 36% of its tech workforce operates remotely, more than any other U.S. city. With 283,000 jobs and an average salary of $135,287, it’s a top pick for those seeking flexibility without cutting pay.

Chicago, Washington D.C., and Boston all offer average tech salaries well above the six-figure mark. Each of these cities also has unique strengths: Chicago has a deep job pool and moderate startup growth; D.C. offers high flexibility with 33.8% of tech roles remote; and Boston remains anchored by strong academic and research institutions.

Austin, San Jose, and Atlanta round out the top 10, all offering average tech salaries above $99,000, though they are behind in job growth and startup activity. 

Again, San Jose, despite being in the heart of Silicon Valley, ranks just ninth among the top U.S. cities for tech jobs, a surprising drop caused by fewer tech events and slower innovation churn compared to its neighbours.

Jessica Bui, spokesperson for TRG Datacenters, said: “As some cities continue to lead with high salaries and a dense concentration of tech events, others demonstrate that strong job growth and thriving start-up ecosystems can be just as impactful. With remote work becoming increasingly common, these cities are providing new opportunities for tech talent, showing that a mix of job opportunities, innovation, and lifestyle can be key drivers for success in the tech industry.”

For tech workers eyeing career moves, there are now more cities than ever where you can earn over $100,000 without needing to live in San Francisco.

In contrast, OpenAI’s GPT-4 generates 25 times more emissions per query, pointing out issues about the environmental cost of advanced AI models.

A recent analysis by TRG Datacenters compared the carbon footprints of various AI models, measuring emissions from individual queries based on standard energy grid assumptions. The results reveal a gap in efficiency, with some models demanding far more power than others.

The AI Carbon Footprint Breakdown

AI chatbots differ widely in how much energy they consume during inference. Here’s how they rank in terms of CO2 emissions per query:

• Grok AI – 0.17g
• Google Gemini – 1.6g
• LLaMA (Meta AI) – 3.2g
• Claude AI – 3.5g
• Perplexity AI – 4g
• ChatGPT (GPT-4) – 4.32g

Grok AI’s low emissions result from its simplified computational design, which reduces power usage while maintaining performance. In practical terms, a single query on Grok produces the same emissions as a basic Google search, making it the greenest option in the market.

Meanwhile, Google Gemini ranks second, emitting 1.6g of CO2 per query. Google’s heavy investment in renewable energy and custom AI hardware helps curb its carbon footprint, though it still lags behind Grok.

Meta’s LLaMA model follows at 3.2g CO2 per query, benefiting from Meta’s commitment to renewable energy but still consuming twice as much power as Gemini. Claude AI ranks slightly worse, producing 3.5g CO2 per query, with its emphasis on safety and reliability seemingly driving up computational costs.

At the higher end of the spectrum, Perplexity AI (4g CO2 per query) and ChatGPT (4.32g CO2 per query) stand out for their environmental impact. GPT-4, in particular, has the highest carbon footprint among the chatbots studied. 

Its computational intensity, deep learning architecture, and search feature demand massive energy resources, generating emissions equivalent to sending 21 emails or nearly a full phone charge per query.

A spokesperson from TRG Datacenters commented on the findings:

“As AI adoption continues to rise, finding ways to reduce its energy consumption will be key. Some models are already designed to be more efficient, but there is still room for improvement. Advances in hardware, more optimized AI models, and increased use of renewable energy in data centres could help lower emissions over time. AI is here to stay, but balancing innovation with sustainability will be essential in minimizing its environmental impact.”

With AI usage skyrocketing, energy efficiency is becoming a big issue. While Grok AI sets the benchmark for low-carbon performance, larger models like GPT-4 highlight the environmental trade-offs that come with high-powered AI capabilities.

The resilience of AI sustainability will probably depend on hardware improvements, algorithmic optimisations, and increased reliance on green energy. For now, the numbers show that not all AI models are created equal, and some come with a much heavier environmental cost than others.

The study analyzed 30 work-related platforms, considering data on the number of major global outages, the average duration of these outages, and the total number of affected users to calculate a reliability score for each service. 

The focus was solely on major global outages where the platform was completely unavailable for all its functions.

Global Outages: Summary of the Findings

Key Highlights

Monday.com was the platform with the most frequent outages, experiencing 10 major crashes last year — almost once every month. These outages lasted an average of 45 minutes and impacted 225,000 users, resulting in the lowest reliability score of 25.5.

Gmail and Microsoft Teams each experienced six major crashes, but the impact was largely different. Gmail’s crashes affected over 1.8 billion users, over 20% of the world’s population, with an average outage duration of 3.5 hours. 

Microsoft Teams, however, had the longest outages in the study, with each lasting over five hours on average, affecting 280 million users. Despite this, Gmail maintained a higher reliability score of 55.0 compared to Microsoft Teams’ 20.7.

Slack and Outlook both had five crashes, but like Microsoft Teams, Outlook’s outages lasted an average of five hours, impacting 400 million users. In contrast, Slack’s crashes were much shorter — just 45 minutes on average — affecting a smaller user base of 18 million, leading to a significantly higher reliability score of 45.8.

Other platforms like Google Drive, GitHub, ClickUp, Salesforce, Zoom, and Notion each experienced four major outages over the last year. Google Drive, despite its widespread usage affecting over a billion users, managed to maintain the highest reliability score in the study at 61.7. The platform’s outages were shorter, averaging 1.5 hours.

Zoom had longer outages, averaging 2.5 hours, impacting 300 million users, resulting in a reliability score of 44. Notion‘s outages were slightly shorter, averaging 2 hours and affecting 30 million users, earning it a reliability score of 42.5.

A spokesperson for TRG Datacenters commented on the importance of these findings:

“The reliability of a platform is easy to maintain when the user base is relatively small. However, the game changes when a single mistake can affect millions or even billions of users, even if the downtime is as short as 30 minutes. Reliability becomes crucial when enterprises depend on these platforms to deliver results. It’s also important to consider that these platforms serve users across different time zones, so not all users may be impacted equally by an outage.”

This study revealed the challenges faced by widely used platforms in maintaining reliable services, especially as their user bases grow. Both businesses and individuals continue to rely heavily on these platforms for daily operations, hence the importance of minimizing downtime and speedily addressing outages cannot be overstated.