科技巨頭
2026年06月13日
Big Challenges Ahead For Meta AI Chief Alexandr Wang After A Rocky First Year
CNBC Television
YouTube
🤖 AI 重點摘要
- Meta 在 AI 投資一年後推出首個自研模型，仍落後於 OpenAI、Anthropic、Google。
- 高層人才流失、裁員與內部矛盾削弱公司士氣。
- Meta 股價下跌 19%，加大了王亞連與 Zuckerberg 的壓力。
- 亟需在廣告之外找到真實收入來源，證明此筆投資的價值。
📝 雙語字幕
▶️
暫停
🔄
1x
It's been one year since
Mark Zuckerberg appointed Alex Wang, Meta's chief AI
officer, as part of a $14 billion
acquihire as Meta invested
已经过去一年，马克·扎克伯格（Mark Zuckerberg）任命 Alex Wang 为 Meta 首席 AI 首席，作为 Meta 投资 140 亿美元的“收购兼招聘”（acquihire）。
in Wang's company,
Scale AI.
在 Wang 的公司 Scale AI。
With Meta shares down 19% in
the past year, far underperforming the
Nasdaq in that period, Meta and Wang's AI direction
is under more scrutiny than
在过去一年里，Meta 股价下跌 19%，在该期间远低于 Nasdaq，Meta 以及 Wang 的 AI 方向正比以往更受审视。
ever.
以往。
The biggest change in
Meta's AI approach from Wang's Meta
Superintelligence Labs has been a shift away from open
source models,
Meta AI 方法中，王实验室的最大变化是远离开源模型，
which were previously
offered for free to the developer community.
这些模型此前是免费向开发者社区提供的。
But its launch of Llama 4
back in April 2025 was considered a big
disappointment.
但其在 2025 年 4 月推出的 Llama 4 被视为一次重大失望。
Meta is now focused on
proprietary tools, with Wang's lab responsible
for launching the new Muse Spark family of models in
April.
现任 Meta 专注于专有工具，王实验室负责在 4 月推出新的 Muse Spark 系列模型。
Seen as progress for Meta,
the company acknowledges that on many metrics,
its core models are less powerful than cutting edge
AI models from Anthropic and
这被视为 Meta 的进步，公司承认在许多指标上，其核心模型的性能低于来自 Anthropic 以及
others.
其他公司。
But basic
functionality is seen as competitive and impressive
when it comes to energy efficiency.
但其基础功能在能源效率方面被视为具竞争力且令人印象深刻。
In a big first,
Wang indicated Meta eventually wants to make
money from these models by offering paid access to
developers,
在一次重大首发中，王表示 Meta 最终计划通过向开发者提供付费访问来获取这些模型的收入，
while in the meantime,
these AI tools bolster Meta's ad business.
在此期间，这些 AI 工具强化了 Meta 的广告业务。
And in another first,
Meta announced a suite of subscription offerings for
its AI features, charging $8 a month for Meta
One Plus and $20 a month for
在另一次首发中，Meta 推出了一套 AI 功能的订阅套餐，月费为 Meta One Plus 每月 8 美元，
Meta One Premium.
以及 Meta One Premium。
Meta's
first year was also marked by personnel upheaval.
Meta 的第一年还伴随人员动荡。
Wang made some high profile
hires from OpenAI and Anthropic, and brought in
former GitHub CEO Nat Friedman, reportedly at
multimillion dollar salary
王招募了些知名人物，包括来自 OpenAI 和 Anthropic 的员工，并引入前 GitHub CEO Nat Friedman，据说其薪酬数千万美元。
levels.
级别。
And sources say
those high paychecks engendered some resentment,
as Wang's group also lost some high level employees
and laid off about 600
据悉，这些高额薪酬引发了部分不满，因为 Wang 的团队也失去了一些高级员工，并裁员约 600 名员工
workers amid reports of
internal conflict and low morale.
在内部冲突和士气低落的报道中。
Now in Wang's second
year, the pressure is on him to
deliver on multiple fronts.
現在是在王的第二年，壓力迫使他必須在多個面向交付成果。
First, he's promised
progress from Muse Spark, saying they expect upcoming
models to be competitive with the leading models in
the world.
首先，他承諾 Muse Spark 取得進展，說他們預計即將推出的模型將與全球領先的模型相競爭。
Second, he's under pressure to
generate new revenue for Meta's AI efforts,
as the company is expected
其次，他正承受著為 Meta 的 AI 事業產生新收入的壓力，因為該公司預計
to spend as much as $145
billion for AI and data center infrastructure this
year, up from $72.5 billion spent
last year.
今年將花費最多 1450 億美元用於 AI 與數據中心基礎設施，較去年花費的 725 億美元更多。
Investors are watching
closely to see if Meta can convert consumers,
companies and developers to pay for its AI models.
投資者密切關注 Meta 是否能說服消費者、公司與開發者為其 AI 模型付費。
Go to cnbc.com for the full
story.
前往 cnbc.com 閱讀完整故事。