A new gold rush? How AI is transforming San Francisco
Visitors made shadow puppets for AI to identify, used AI to generate songs, asked chatbots questions and faced off with AI in a game in which players tried to draw images that only humans would recognize. A giant robot hand moved around and people peered into a video game chip.
They jotted down their hopes and worries about AI on cards displayed in the museum. Hope: AI will cure cancer. Worry: People will rely on AI to the point they can't think for themselves.
'It sort of breaks down those guardrails, those big walls that people have put up around AI, and allows them to have a conversation with somebody else,' said Doug Thistlewolf, who manages exhibit development at the Exploratorium.
Art. Office Space. Billboards. Protests. The AI craze has intensified in San Francisco, spreading through work and social life in what some have described as a new gold rush. The AI boom, coupled with the election of new Mayor Daniel Lurie, has also infused the city with optimism — tinged with anxiety. Some worry about the city's high cost of living, and whether AI will replace workers as tech layoffs continue.
Read more: Bay Area tech workers thought their jobs were safe. Then the 'golden handcuffs' came off
For years, Silicon Valley has been at the center of innovation with some of the world's valuable tech companies such as Meta, Google, Apple and Nvidia locating their massive headquarters south of San Francisco. AI's rise, though, has shone a bright spotlight on San Francisco, home to multibillion-dollar companies such as OpenAI, Scale AI, Anthropic, Perplexity and Databricks.
AI has long played a big role in consumer technology, helping to recommend social media posts, translate languages and power virtual assistants. But the popularity of OpenAI's ChatGPT — a chatbot that can generate text, images and code — set off a fierce race to propel technology that touches industries from media to healthcare.
Companies are battling it out for talent, offering lucrative compensation to recruit top researchers and leaders, while investments in AI companies have surged.
In the first half of 2025, venture capital funding for AI companies in the San Francisco Metro area surpassed $29 billion — more than double the amount during the same period in 2022, data from PitchBook shows. As of Aug. 5, VC deals for AI startups in the area, which includes San Francisco, Oakland and Fremont, made up 46.6% of funding for U.S. AI companies this year.
Exactly how this frenzy will shape the future of San Francisco, home to cable cars and robotaxis, remains to be seen. Ask ChatGPT what SF will look like in 10 years and it generates an image of the city's skyline with futuristic architecture and flying saucers next to the Golden Gate Bridge.
AI has been a 'bright spot' in the city's economy, helping San Francisco to recover after retailers, office workers and some companies such as X (formerly Twitter) left the downtown area during and after the pandemic as remote work picked up.
'The economic impact is [AI companies] take more office space, they pay more taxes, they hire more people,' said Ted Egan, chief economist of the city and county of San Francisco.
Over the past five years, AI-related companies have leased more than 5 million square feet of San Francisco office space and the amount is projected to grow, according to CBRE, a real estate service and investment firm. The city's office vacancy rate of 35.8% in the first quarter would be cut in half if these companies take up 16 million square feet of office space by 2030.
Read more: There's one bright spot for San Francisco's office space market
San Francisco resident Vijay Karunamurthy has seen the city's boom and bust cycles unfold over the last 25 years while working at startups and tech giants such as Google and Apple.
In 2000, when he moved from Chicago to San Francisco for an engineering job at a data startup, he saw major business such as Pets.com collapse during the dot-com crash. Fueled by social media's popularity, the city's tech sector came roaring back only to take a hit during the COVID-19 pandemic.
Now the city is ascending yet again. Ambitious entrepreneurs, old and new, are advancing powerful artificial intelligence tools that could transform lives.
'That amount of energy being concentrated in San Francisco has just been huge for the city,' said Karunamurthy, 46, the former field chief technology officer at Scale AI, a data-labeling startup. 'It means every single night there's AI events, and if you go to a coffee shop, you'll run into people working on AI.'
Still, there are plenty of AI skeptics. In late July, outside of OpenAI's headquarters in Mission Bay, a small group of protesters including a person dressed up as a robot held up signs that said "AI will kill us all" and "AI steals your work to steal your jobs."
Generative AI's ubiquity has forced educators to rethink what and how they teach students in the classrooms.
Arno Puder, professor and chair of San Francisco State University's computer science department, said generative AI represents a historic 'paradigm shift.'
The longtime San Francisco resident is equally excited, but also a little scared, about how it will affect labor. Over the last two years, he's seen student enrollment in computer science at the university drop amid tech layoffs and generative AI's rise. As coding assistants reshape computer science jobs, the university launched a new undergraduate certificate in generative AI for the fall of 2026.
"Generative AI is a different beast," Puder said. "That does make me worry a little bit, but if you ask me for a prediction on what services or what the world's going to look like in a few years from now, I don't know."
AI's rise has inspired the creation of new spaces throughout San Francisco where people can discuss technology's benefits and risks.
Thistlewolf said creating the AI exhibit at the Exploratorium involved talking to workers and researchers from tech companies and universities. The exhibit, which runs through mid-September, took roughly a year and half to develop.
Backed by Anthropic, the San Francisco company that developed the AI chatbot Claude, the exhibit aims to educate people about AI but doesn't shy away from the debate surrounding technology.
San Francisco resident Martha Chesley, 77, came to the exhibit with her grandchildren. Living in San Francisco for 50 years, Chesley sees potential benefits from AI companies coming to the city.
'If it brings people and money, it's good for the city because right now we have a lot of closed storefronts,' she said. 'Maybe there would be more money also for housing being built.'
Throughout the city, AI startups are broadcasting their mission loudly on billboards and ads displayed at bus stops and train stations. Messages include 'Stop Hiring Humans. To Write Cold Emails' and 'Droids ship software while you touch grass.'
AI ads could also be spotted in the Mission district, a neighborhood deeply rooted in Latino culture and history. The area, filled with popular taquerias, colorful murals and a park with a view of the downtown skyline, has struggled with homelessness like other parts of the city.
At a bus stop on 16th Street, an ad from AI startup Outset struck a positive tone: 'Listen to humans. Don't replace them.'
Founded in downtown San Francisco in 2022, Outset created an AI interviewer so researchers could quickly gather feedback from more people to better understand customer needs and improve products.
The company's 36-year-old chief executive, Aaron Cannon, said before the rise of ChatGPT, he and his co-founder experimented with AI systems that can generate and understand human language and saw its potential.
'I don't think either of us could have told you it was going to absolutely take over the world,' he said. The San Francisco resident said the city's talent pool also makes it an attractive location for startups. He declined to disclose its finances but said the company, which employs 15 and counts Microsoft among its clients, is "growing fast."
Throughout San Francisco, founders and real estate companies have dubbed certain areas as AI hubs.
Hayes Valley, a neighborhood with Victorian houses, boutique shops and trendy restaurants, bears the nickname 'Cerebral Valley,' a nod to the hacker houses and AI communities that popped up in the area.
Jamestown, a real estate and investment company, markets the Northern Waterfront an emerging AI hub after leasing more than 43,000 square feet of office space to AI companies. Some of the startups work on AI loan servicing or AI-powered lip syncing technology.
Located near public transportation, water and greenery, the fresh air and serene nature of the area has attracted AI entrepreneurs that want to collaborate in person, said Michael Phillips, principal and chairman of Jamestown.
'If you're working on these fast to market, highly competitive products," he said, "you really need to be together.'
Sign up for our Wide Shot newsletter to get the latest entertainment business news, analysis and insights.
This story originally appeared in Los Angeles Times.
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Newsweek
13 minutes ago
- Newsweek
Ford's New Assembly Line Design is Key to Lowering Vehicle Costs
Based on facts, either observed and verified firsthand by the reporter, or reported and verified from knowledgeable sources. Ford tore up the moving assembly line concept and designed a better one, the company's CEO Jim Farley explained to a crowd gathered at Louisville Assembly Plant in Kentucky for Ford's announcement that a new, lower-cost electric truck is coming to market in 2027. "One of the first things we did was bring in manufacturing expertise directly from our teams in Dearborn. That was incredibly important [it] set the baseline of what the architecture of the products would be off the platform. [It is a] marriage of both worlds, in terms of the expertise of 122-year-old company with massive capabilities and supply chain and industrialization in general," Alan Clarke, Ford's executive director of advanced EV development, told Newsweek. "[It] was really important to not only engage early, but also ensure that the ideas were able to fail fast. That's a startup mentality. When you're moving quickly, you want to really quickly determine: Is this actually more efficient? Does it actually save us money? Do we still deliver what we need to the customer in order to effectively make the product?" he said. The formula for the new assembly line structure, the Ford Universal EV Production System, starts with the new Ford Universal EV Platform. That platform underpins future battery-electric vehicles and is the result of a California-based skunkworks operation that was designed and enabled to break Ford's timeline, agility and product development norms. F-150 Lightning trucks coming down an assembly line. F-150 Lightning trucks coming down an assembly line. Ford Motor Company Instead of a traditional snaking assembly line structure, Ford's Louisville plant, which expanded 52,000 square feet as part of the effort, will feature an "assembly tree." That tree will include three sub-assemblies that run down their own separate assembly lines in unison, then join together at the end. "There's been a lot of work in the last three years on AI, on plants, and we're doing that and putting that into the system too. You've seen some of that already, but then we've also taken to this new platform, what we're able to do and add to that. So again, it's a win, win on both sides," a Ford spokesperson said. As part of the vehicle building shift, single-piece aluminum unicastings will replace dozens of smaller parts. This will allow the front and rear of the vehicle to be assembled separately. Additionally, less parts means fewer opportunities for failures that result in recalls, a business element that Ford has continuously struggled with during Farley's tenure, costing the company billions. "Elegant, simple designs are ultimately what enable low-cost: deletion of parts, deletion of process.... There's a lot of innovation of powertrain, in how we put the whole product together," the spokesperson said. Under the new assembly plan, Ford's assembly line workers on the third branch of the line put together the battery, seats, consoles and carpeting. Parts travel down the assembly tree to operators, further simplifying operations, in kit form with all fasteners, scanners and power tools included, in the order in which they are intended to be used. The team has further optimized the assembly line and order of operations for building the vehicle to improve the ergonomics for employees, reducing twisting, reaching and bending. Ford estimates that this new way of assembling vehicles will be 40 percent faster than the current assembly time of the Ford Escape. Some of that time, Ford says, will be yielded back to the company for other tasks, with an ultimate net speed improvement of 15 percent. It takes 20 hours to build one Ford Escape today.
CNBC
13 minutes ago
- CNBC
A new piece of Wall Street research prompted us to change our rating on Salesforce
For the better part of the last decade, we've heard that software is eating the world. But now it seems AI is eating software. That's the call analysts at Melius Research are once again making Monday as they downgraded one-time market darling Adobe to a sell rating. The attention-grabbing note highlights the far-reaching ripple effects that artificial intelligence is having across the stock market — and investors must be attune. Club name Salesforce is in the crosshairs of the shifting landscape, and we cannot ignore that fact. It's a major reason why the stock and that of enterprise software peers Atlassian and Workday have struggled in 2025. In Melius' view, the rise of AI is to software what the rise of cloud computing was to on-premise data centers. Historically, companies would maintain their own data centers in-house, requiring massive investments in hardware and teams of engineers there to maintain and troubleshoot any issues. Today, however, companies can outsource this simply by hopping on the cloud computing bandwagon, utilizing the likes of Amazon Web Services, Microsoft's Azure and Alphabet's Google Cloud. AWS became the first modern cloud service to launch in 2006. The rise of the cloud was bad news for the stocks of on-premise hardware providers such as Dell , Hewlett-Packard and IBM . As this shift to cloud computing took hold, Melius noted that these companies all saw a steep contraction in their price-to-earnings multiples, a classic sign that investors believe slower growth is ahead. "Just when you thought the coast was clear, their PE multiples kept going lower and lower — from the 20s to mid-single digits," Melius wrote Monday. Club name Microsoft was also a key provider of software for on-premise servers. However, Microsoft did adjust its strategy with the launch of cloud service Azure. The cloud was a priority for CEO Satya Nadella upon taking over in early 2014, and it's paid off handsomely for investors — just take a look at the stock chart versus the S & P 500. MSFT .SPX mountain 2013-12-31 Microsoft's stock performance versus the S & P 500 since the start of 2014. Microsoft would have no doubt been on that list of on-premise fallen angels — even if not a hardware play like Hewlett-Packard and Dell — had they not jumped on cloud computing and instead kept their fate tied to the on-premise hardware providers. Azure is the No. 2 cloud by revenue behind AWS and is seeing growth reaccelerate thanks to AI. The lesson from Microsoft: If you're business model is being disrupted, you can still adapt, survive and thrive. However, if you're getting disrupted and fail to adapt, then the valuation that investors are willing to pay for your stock will reflect that failure. Instead, the investors will move on to owning the disruptors in question, which, in the present moment, are the AI players pushing into the software-as-a-service space. Often abbreviated as SaaS, this business model was all the rage not too long ago itself. Instead of companies paying for a perpetual license to use a specific version of software, SaaS companies sold their product on a subscription basis, bringing in a more predictable recurring revenue stream that investors were willing to assign a higher multiple. It often involves a "seat model," where companies pay based on the number of employees using the software. Melius believes it's under real threat. "We think companies are increasingly understanding that AI tools can help cut expensive knowledge workers with those costly SaaS seats. In fact, opex is going down as a percent of sales across tech leaders since the AI boom started. One of the ways they are doing it is reducing workers who typically are the biggest consumers of SaaS ... and Adobe could be impacted as other companies catch on," the Melius analysts wrote in their downgrade. They added, "We think there will be FOMO across all industries to cut costs and get their stocks up - SaaS is the casualty as AI adoption accelerates." To be sure, while this is a potential concern for Salesforce, plenty of other tech names in the portfolio stand to benefit. Among them are Microsoft and Amazon because their cloud services are enabling others to leverage AI. Indeed, the Melius analysts themselves contend: "While SaaS should be avoided, 'Software' companies with clouds [such as Microsoft and Oracle] are continuing to see acceleration in demand." "Another area fueling demand is the acceleration of AI agents that do the work of SaaS and fuel demand for more compute," Melius added. Agentic AI is a type of AI system that can complete certain tasks without human intervention, and it's become a big area of investment and focus in the AI race — including for Salesforce. In that way, Salesforce occupies both sides of a tricky situation. It's both a SaaS company and a company that's recognized how transformative AI can be. That's why we were torn as we read Melius' research Monday. On the one hand, we think Melius makes some pretty good points. After all, we are seeing companies do more with less thanks to generative and agentic AI. And given that SaaS is generally based on the seat-based licensing model, if Salesforce's customers have fewer employees because their remaining ones are using AI to be more productive, that translates to less demand for software seats — all else being equal. We also do believe AI is a once-in-a lifetime technology that will have implications far wider reaching than the shift to cloud did over the past decade. On the other hand, Salesforce in particular remains a critical tool for companies across industries and a best-in-class provider when it comes to customer relationship management (CRM) software. Salesforce also isn't standing idly by. Instead, the company is working to build out its agentic AI offering, appropriately dubbed Agentforce, so that it can help clients move more quickly into the AI era, while also building out a new AI-oriented revenue stream that can hopefully help extract more revenue from those seats at customers that do remain. Unlike Adobe, it's also important to keep in mind that Salesforce is targeted almost exclusively toward enterprise, where the customer cohort tends to be more sticky and a bit slower to change things up, lest they risk a misstep that hurts sales. While Salesforce isn't alone in offering tools to build Agents — Microsoft, which competes with Salesforce in the CRM arena, also has them — the company clearly has a strategy. In the end, we aren't ready to bail on Salesforce just yet, though we understand the market's concern with the SaaS space. Salesforce is still a great house, but the neighborhood is deteriorating. And as any good real estate agent will tell you, it's location that matters above all. As a result, we're downgrading shares of Salesforce to a hold-equivalent 2 rating as we look to better understand to what extent Salesforce is being disrupted, or if can indeed adapt to the new world. While the stock is already down a lot this year, and we believe in the management team, it's the proof in the numbers that we need to see before getting more optimistic. (Jim Cramer's Charitable Trust is long CRM. See here for a full list of the stocks.) As a subscriber to the CNBC Investing Club with Jim Cramer, you will receive a trade alert before Jim makes a trade. Jim waits 45 minutes after sending a trade alert before buying or selling a stock in his charitable trust's portfolio. If Jim has talked about a stock on CNBC TV, he waits 72 hours after issuing the trade alert before executing the trade. THE ABOVE INVESTING CLUB INFORMATION IS SUBJECT TO OUR TERMS AND CONDITIONS AND PRIVACY POLICY , TOGETHER WITH OUR DISCLAIMER . NO FIDUCIARY OBLIGATION OR DUTY EXISTS, OR IS CREATED, BY VIRTUE OF YOUR RECEIPT OF ANY INFORMATION PROVIDED IN CONNECTION WITH THE INVESTING CLUB. NO SPECIFIC OUTCOME OR PROFIT IS GUARANTEED.
CNN
13 minutes ago
- CNN
Massive revisions shook the jobs report. Tuesday's inflation data could get cloudy, too
The next batch of inflation data from the Bureau of Labor Statistics was already shaping up to be a high-profile affair due to the expected impact of President Donald Trump's hefty tariffs. But after Trump fired the agency's top statistician, Tuesday's report now comes with some other, unexpected baggage. July's Consumer Price Index is set to be released at 8:30 a.m. ET on Tuesday, just 11 days after Trump's unprecedented firing of the head of the BLS, following a lackluster jobs report that contained substantial revisions to prior months of employment data. The president baselessly claimed that BLS Commissioner Erika McEntarfer 'rigged' the data (an allegation that Trump's former BLS head and a cadre of statisticians and economists resoundingly disputed). Moreover, the data that feeds into the CPI could make the closely watched inflation gauge subject to greater revisions, in part due to Trump administration's budget and staffing cuts. 'There's going to be a larger margin of error with an expectation of rolling revisions to the CPI data in a way that is somewhat similar to what we see in the jobs data and the durable goods data,' Joe Brusuelas, chief economist at RSM US, told CNN. The BLS, one of many federal agencies subject to the Department of Government Efficiency's blunt axe, has increasingly cut back on the data collection that feeds into the critical pricing gauge. Since April, the BLS stopped collecting data in three not-so-small metro areas (Buffalo, New York; Lincoln, Nebraska; and Provo, Utah) and further leaned on an imputation process used to estimate prices by means other than direct, local observations (such as making price comparisons on a regional or national basis instead). The BLS said in early June that the tri-city collection reductions (which amounted to roughly 3.6% of the CPI's overall sample) were expected to have 'minimal' impact on the overall index but increase the volatility of geographic- or item-specific price indexes. Two weeks ago, the agency shared a statistical analysis of those cutbacks using a 77-month study period. It showed that the annual inflation reading was one-tenth of a percentage point higher in 18% of those months and one-tenth of a percentage point lower in 14% of those months. In that same notice, however, BLS also disclosed that it made a 15% wholesale reduction in collections across the other 72 coverage areas. 'Collection suspension affects both the Commodity and Services Pricing survey and the Housing survey,' the BLS wrote. 'As a result, the number of collected prices and the number of collected rents used to calculate the CPI has temporarily been reduced.' The BLS did not share any statistical analysis of those wider cuts nor provide enough information for others to estimate that, Alan Detmeister, a UBS senior economist who previously helmed the Federal Reserve's Wages and Prices section, told CNN. 'The 15% of the remaining sample is four times larger than the number of observations dropped than those three cities,' he said. 'And if those three cities already made it round differently a third of the time. You're probably talking a tenth or so on overall inflation. Not huge, but it's something.' 'It's definitely degrading the quality of the inflation statistics,' he said. While the reductions in price observations and increase in imputations could make the CPI data — especially on a monthly basis — less sharp, it's not expected to cause similar large variations as seen in the survey-driven jobs report, which is reliant on an increasingly shrinking sample of respondents, RSM's Brusuelas said. 'It's going to bore us to sleep watching the back revisions in the CPI and [Personal Consumption Expenditures] data,' he said. The issues of greater importance involving the CPI data is what Tuesday's numbers say about the trajectory of inflation and any coinciding pushback from the Trump administration on any report that shows tariffs are driving prices higher, Brusuelas said. For any presidential administration, that's 'tantamount to engaging in self-harm, attempting to tell the public that prices aren't rising, when they can clearly feel it in their own bottom lines and reduced purchasing power of their paychecks,' he said. 'You saw what happened when the Biden administration tried to play down price increases as inflation growth was slowing. It did not resonate.' In June, CPI rose to 2.7%, its highest level in four months, as price increases — including those from tariffs — packed a bigger punch. July's CPI is expected to tell a similar tale: Tariffs are causing higher prices on a wider swath of goods while falling gas prices and tepid consumer demand is keeping a lid on some services prices. All in all, the index is expected to rise 0.2% for the month and tick up to 2.8% on an annual basis, according to FactSet consensus estimates. The July CPI will show 'further evidence of tariff-induced inflation eroding the purchasing power of consumer paychecks,' Brusuelas said. 'The thing to understand is this is not going to happen with a bang but more of a slow deterioration in purchasing power.'
